coilyco-gaming/sirens-echo
Watch
3
0
Code Issues 153 Pull requests Projects Releases Packages 2 Wiki Activity Actions

Sirens Deep: compose a real Agent Compose identity, selected by one deploy string #98

New issue
Closed
opened 2026-08-11 04:29:45 +00:00 by coilyco-ops · 6 comments
coilyco-ops commented 2026-08-11 04:29:45 +00:00
Member
Copy link

Outcome

Sirens Deep gets a genuine Agent Compose identity, the full role and personality meld, flippable between roles by changing one string in the deploy repository. Sirens Echo stays void of personality exactly as it is today.

Design agreed with Kai on 2026-08-10. This issue is the durable record.

The split

  • Sirens Echo keeps its neutral profile and Sirens Discord knowledge. <aos-community-bundle> and personality meld stay forbidden in its rendered prompt.
  • Sirens Deep composes a role bundle. Those same markers become required, so a bundle that silently fails to compose is a startup failure rather than a quietly neutral bot.

ValidateSystemPrompt already enforces the forbidden direction. The change inverts it per profile rather than adding a new mechanism.

Source selection: aos public catalog plus an explicit aosk allowlist

The aos .agents/composed/ catalog is public-safe as a whole. The aosk one is not uniformly, so the compose request names an allowlist and excludes everything else by default.

In, all public-safe and mirrored on the public website:

  • writing-kai-voice (now Coilyco house style)
  • personal-preference-colors, -animals, -games, -shows, -anime, -books, -movies, -fabrication

Out, and the reason matters:

  • kai-career, kai-job-search, kai-grill-me, kai-collaboration - private context.
  • personal-preference-social - an organization can own a favorite color; it cannot own a person's social accounts. The reframe in coilyco-bridge/agentic-os-kai#853 works for tastes and for style, and does not work for biographical facts about a member. That distinction is the general rule for future additions.
  • kai-engineering-voice - portable after the de-Kapwinging in #853, but it is about producing code reviews and eng-channel posts, which is not what this agent does.
  • kai-design-language - public-safe but marked low-context: required and it is art direction.

A pre-commit check should fail on any source added outside the allowlist, so widening the surface is a reviewed act.

Shape

agent/compose-request.kdl names the catalog, the personality library, and the allowlist. The image build runs agent-compose bundle materialize once per role and bakes all eight. Deploy sets one variable:

SIRENS_DEEP_ROLE=director

Nothing is fetched upward at runtime, which keeps the config-placement rule intact: this repository authors the schema and the composition, coilyco-bridge/deploy owns the value.

Why this is reviewable

agent/rendered/*.prompt.txt already snapshots the assembled prompt and a pre-commit hook fails on drift. Extend it to emit one snapshot per role, so flipping SIRENS_DEEP_ROLE produces a reviewable diff of exactly what that persona tells the model. Current sizes for reference: Echo 5717 bytes from two policy roots, Deep 3091 from one. A role bundle will move that materially, and the byte count is the cheapest early warning.

The favorite-colour question, resolved

Agent Compose gives each seat its own accent (#b39258 engineer, #cb7471 director). The composed favorites give Coilyco purple and black. These do not collide once the favorites belong to the organization: the seat colour is the agent's own, the palette is the organization's. No post-processing needed.

Never impersonate

The agent shares house taste and house style. It never claims to be a specific person and never answers as one. People in a guild the operator does not moderate cannot be left unsure whether they are talking to a real human.

Prerequisites

  • internal/community/skillpack.go reads COMPOSED.md - done, landed in #93.
  • The composed sources are reframed onto the organization - done, coilyco-bridge/agentic-os-kai#853.
  • agent-compose must exist inside forgejo.coilysiren.me/coilyco-flight-deck/agentic-os:release. Unverified. The whole build-time bundling approach depends on it, and if it is absent the bundle has to be produced elsewhere and copied in.

Complete when

  • A tracked compose request names the catalog and the allowlist, and a check rejects sources outside it.
  • The image carries one verified bundle per role.
  • SIRENS_DEEP_ROLE selects the role at deploy time with no rebuild.
  • Sirens Deep's rendered prompt contains the role and personality surface; Sirens Echo's still does not, and both are asserted.
  • One prompt snapshot per role is tracked and gated.
  • The agent shares taste and style without claiming to be a person.

Related

  • coilyco-bridge/agentic-os-kai#855 - the pronoun rail regression introduced by the scrub, which this agent will inherit.
  • #87 - the deployed agent misgendering a member, which the composed sources are meant to prevent.
  • #76 - the August 19 gap, of which the persona is one item.
## Outcome Sirens Deep gets a genuine Agent Compose identity, the full role and personality meld, flippable between roles by changing one string in the deploy repository. Sirens Echo stays void of personality exactly as it is today. Design agreed with Kai on 2026-08-10. This issue is the durable record. ## The split * **Sirens Echo** keeps its neutral profile and Sirens Discord knowledge. `<aos-community-bundle>` and `personality meld` stay **forbidden** in its rendered prompt. * **Sirens Deep** composes a role bundle. Those same markers become **required**, so a bundle that silently fails to compose is a startup failure rather than a quietly neutral bot. `ValidateSystemPrompt` already enforces the forbidden direction. The change inverts it per profile rather than adding a new mechanism. ## Source selection: aos public catalog plus an explicit aosk allowlist The aos `.agents/composed/` catalog is public-safe as a whole. The aosk one is not uniformly, so the compose request names an allowlist and excludes everything else by default. **In**, all public-safe and mirrored on the public website: * `writing-kai-voice` (now Coilyco house style) * `personal-preference-colors`, `-animals`, `-games`, `-shows`, `-anime`, `-books`, `-movies`, `-fabrication` **Out**, and the reason matters: * `kai-career`, `kai-job-search`, `kai-grill-me`, `kai-collaboration` - private context. * `personal-preference-social` - **an organization can own a favorite color; it cannot own a person's social accounts.** The reframe in `coilyco-bridge/agentic-os-kai#853` works for tastes and for style, and does not work for biographical facts about a member. That distinction is the general rule for future additions. * `kai-engineering-voice` - portable after the de-Kapwinging in #853, but it is about producing code reviews and eng-channel posts, which is not what this agent does. * `kai-design-language` - public-safe but marked `low-context: required` and it is art direction. A pre-commit check should fail on any source added outside the allowlist, so widening the surface is a reviewed act. ## Shape `agent/compose-request.kdl` names the catalog, the personality library, and the allowlist. The image build runs `agent-compose bundle materialize` once per role and bakes all eight. Deploy sets one variable: ``` SIRENS_DEEP_ROLE=director ``` Nothing is fetched upward at runtime, which keeps the config-placement rule intact: this repository authors the schema and the composition, `coilyco-bridge/deploy` owns the value. ## Why this is reviewable `agent/rendered/*.prompt.txt` already snapshots the assembled prompt and a pre-commit hook fails on drift. Extend it to emit one snapshot per role, so flipping `SIRENS_DEEP_ROLE` produces a reviewable diff of exactly what that persona tells the model. Current sizes for reference: Echo 5717 bytes from two policy roots, Deep 3091 from one. A role bundle will move that materially, and the byte count is the cheapest early warning. ## The favorite-colour question, resolved Agent Compose gives each seat its own accent (`#b39258` engineer, `#cb7471` director). The composed favorites give Coilyco purple and black. These do not collide once the favorites belong to the organization: the seat colour is the agent's own, the palette is the organization's. No post-processing needed. ## Never impersonate The agent shares house taste and house style. It never claims to be a specific person and never answers as one. People in a guild the operator does not moderate cannot be left unsure whether they are talking to a real human. ## Prerequisites * `internal/community/skillpack.go` reads `COMPOSED.md` - **done**, landed in #93. * The composed sources are reframed onto the organization - **done**, `coilyco-bridge/agentic-os-kai#853`. * `agent-compose` must exist inside `forgejo.coilysiren.me/coilyco-flight-deck/agentic-os:release`. **Unverified.** The whole build-time bundling approach depends on it, and if it is absent the bundle has to be produced elsewhere and copied in. ## Complete when * A tracked compose request names the catalog and the allowlist, and a check rejects sources outside it. * The image carries one verified bundle per role. * `SIRENS_DEEP_ROLE` selects the role at deploy time with no rebuild. * Sirens Deep's rendered prompt contains the role and personality surface; Sirens Echo's still does not, and both are asserted. * One prompt snapshot per role is tracked and gated. * The agent shares taste and style without claiming to be a person. ## Related * `coilyco-bridge/agentic-os-kai#855` - the pronoun rail regression introduced by the scrub, which this agent will inherit. * #87 - the deployed agent misgendering a member, which the composed sources are meant to prevent. * #76 - the August 19 gap, of which the persona is one item.
coilyco-ops commented 2026-08-11 23:17:36 +00:00
Author
Member
Copy link

Prerequisite resolved: agent-compose is in the release image

The issue lists this as Unverified and says the whole build-time bundling approach depends on it. It holds.

coilyco-flight-deck/agentic-os docker/dev-base/install-common.sh:118 installs the binary to /usr/local/bin/agent-compose and symlinks acompose; verify-common.sh runs agent-compose version and agent-compose roster as a build check. This repository's Dockerfile already builds FROM forgejo.coilysiren.me/coilyco-flight-deck/agentic-os:release in both stages, so the tool is present where bundling would run.

The CLI surface matches the plan: agent-compose compose [request.kdl] --out --target and agent-compose bundle materialize --role --harness --out, plus verify and project --scope home for the staged-home handoff. Local version v2.19.0.

Blocker: a compose request cannot narrow what a source contributes

The plan's core mechanism does not exist. It says the compose request "names an allowlist and excludes everything else by default". Agent Compose does not work that way.

Selection is owned by each source repository's own .agents/roles.kdl, not by the requesting repository. Per docs/kdl-contracts.md, a request names role, delivery, model-tier, and source entries. It has no vocabulary for restricting a source's catalogue. agent/compose/roles.kdl in this repository is therefore inert as a filter: it would only bind this repository's own .agents/composed/ catalogue, and this repository has none.

Reproduction

Composed role "creator" against both catalogues in an isolated HOME and PROJECTS_ROOT:

compose {
    role "creator"
    delivery "native-skills"
    source "aos-public" root="projects/coilyco-flight-deck/agentic-os" required=#true
    source "aos-kai" root="projects/coilyco-bridge/agentic-os-kai" required=#true
}

Every source the issue puts Out was selected anyway, because coilyco-bridge/agentic-os-kai's own creator role binds them:

  • kai-career, kai-job-search - named as private context
  • kai-linkedin-voice, kai-linkedin-video - a member's personal channel
  • personal-preference-social - the exact case the favorite-colour rule was written to exclude
  • kai-engineering-voice - named as out of scope for this agent

It also pulled the full ordinary catalogue, including repo-lore, repo-inbox, repo-voice-corpus, repo-deploy, repo-infrastructure, repo-agentic-os-xxx, and every mcp-tools-* pointer. Removing operating_context from the host config changed nothing: same bundle hash, same selection.

aos-public alone stays public-safe and does deliver the wanted writing-* craft skills, but it drops personal-preference-* and writing-kai-voice, which are the two things the issue wants from the aosk catalogue. It also carries unrelated ordinary skills (tooling-gpg-ssm, tooling-tailscale, tooling-warp, tooling-zsh, kai-tech-prefs).

What this means

The allowlist has to move to wherever the binding actually happens, which is the source repository. The choice of where is a design decision, not a mechanical fix, so the wiring stops here pending it. The remaining items in Complete when are all downstream of it: the Dockerfile compose stage, SIRENS_DEEP_ROLE, the validator inversion, and the per-role snapshots.

The prerequisite bullet can be marked done regardless.

## Prerequisite resolved: agent-compose is in the release image The issue lists this as **Unverified** and says the whole build-time bundling approach depends on it. It holds. `coilyco-flight-deck/agentic-os` `docker/dev-base/install-common.sh:118` installs the binary to `/usr/local/bin/agent-compose` and symlinks `acompose`; `verify-common.sh` runs `agent-compose version` and `agent-compose roster` as a build check. This repository's `Dockerfile` already builds `FROM forgejo.coilysiren.me/coilyco-flight-deck/agentic-os:release` in both stages, so the tool is present where bundling would run. The CLI surface matches the plan: `agent-compose compose [request.kdl] --out --target` and `agent-compose bundle materialize --role --harness --out`, plus `verify` and `project --scope home` for the staged-home handoff. Local version v2.19.0. ## Blocker: a compose request cannot narrow what a source contributes The plan's core mechanism does not exist. It says the compose request "names an allowlist and excludes everything else by default". Agent Compose does not work that way. Selection is owned by **each source repository's own** `.agents/roles.kdl`, not by the requesting repository. Per [`docs/kdl-contracts.md`](https://forgejo.coilysiren.me/coilyco-flight-deck/agent-compose/src/branch/main/docs/kdl-contracts.md), a request names `role`, `delivery`, `model-tier`, and `source` entries. It has no vocabulary for restricting a source's catalogue. `agent/compose/roles.kdl` in this repository is therefore inert as a filter: it would only bind this repository's own `.agents/composed/` catalogue, and this repository has none. ### Reproduction Composed `role "creator"` against both catalogues in an isolated `HOME` and `PROJECTS_ROOT`: ```kdl compose { role "creator" delivery "native-skills" source "aos-public" root="projects/coilyco-flight-deck/agentic-os" required=#true source "aos-kai" root="projects/coilyco-bridge/agentic-os-kai" required=#true } ``` Every source the issue puts **Out** was selected anyway, because `coilyco-bridge/agentic-os-kai`'s own creator role binds them: * `kai-career`, `kai-job-search` - named as private context * `kai-linkedin-voice`, `kai-linkedin-video` - a member's personal channel * `personal-preference-social` - the exact case the favorite-colour rule was written to exclude * `kai-engineering-voice` - named as out of scope for this agent It also pulled the full ordinary catalogue, including `repo-lore`, `repo-inbox`, `repo-voice-corpus`, `repo-deploy`, `repo-infrastructure`, `repo-agentic-os-xxx`, and every `mcp-tools-*` pointer. Removing `operating_context` from the host config changed nothing: same bundle hash, same selection. `aos-public` alone stays public-safe and does deliver the wanted `writing-*` craft skills, but it drops `personal-preference-*` and `writing-kai-voice`, which are the two things the issue wants from the aosk catalogue. It also carries unrelated ordinary skills (`tooling-gpg-ssm`, `tooling-tailscale`, `tooling-warp`, `tooling-zsh`, `kai-tech-prefs`). ### What this means The allowlist has to move to wherever the binding actually happens, which is the source repository. The choice of where is a design decision, not a mechanical fix, so the wiring stops here pending it. The remaining items in **Complete when** are all downstream of it: the Dockerfile compose stage, `SIRENS_DEEP_ROLE`, the validator inversion, and the per-role snapshots. The prerequisite bullet can be marked done regardless.
coilyco-ops commented 2026-08-11 23:22:00 +00:00
Author
Member
Copy link

Correction: the allowlist mechanism already exists

My previous comment claimed a compose request cannot narrow a source. That is wrong, and I am correcting it before anything gets built on it.

A request's source node takes exactly one of root or declaration (internal/schema/schema.go:271-295). Everything above used root=, which is the "take the source's whole catalogue as its own roles.kdl dictates" form. The declaration= form is the other half: the consumer names a .kdl file that enumerates precisely the skills it admits, by explicit path (schema.go:918-960). That is a consumer-side allowlist, and it is already in the grammar and documented in docs/kdl-contracts.md.

So the plan's mechanism is real. It just is not spelled the way this issue assumed, and it needs no new file convention or agent-compose change.

Verified

compose {
    role "creator"
    delivery "native-skills"
    source "aos-public" declaration="staged/aos-public/aos-public.kdl" required=#true
    source "aos-kai" declaration="staged/aos-kai/aos-kai.kdl" required=#true
}

Result, agent-compose verify: bundle verified: 15 skills // 18 files

  • roster:core - role-creator, personality-editorial, personality-nurturing, personality-warm
  • aos-public - tooling-discord-community-host, tooling-customer-success-signal-routing, tooling-customer-success-trust-repair, writing-social-cultural-reading, writing-social-editorial-loop, writing-social-trust-boundaries, writing-voice-adaptation
  • aos-kai - writing-kai-voice, personal-preference-colors, personal-preference-games, personal-preference-animals

Nothing outside the named set. No kai-career, no kai-job-search, no kai-linkedin-*, no personal-preference-social, no repo-* pointers, no mcp-tools-*. Compare the root= run in the previous comment, which pulled all of those.

What carries over from the previous comment

Two findings stand.

The roster is supplied by agent-compose itself, not by a source. role-creator and the three personalities arrive even when both sources are bounded declarations. Sirens Deep gets its full identity and meld regardless of how narrow the catalogue selection is.

root= on a source is unsafe here. With root=, the source repository's own roles.kdl decides, and coilyco-bridge/agentic-os-kai's creator role deliberately binds Kai's career, job-search, and LinkedIn context because that role serves Kai. This repository must use declaration= for both sources, never root=.

Consequences for the plan

  • agent/compose/roles.kdl is the wrong document. It is a provider-side graph for a catalogue this repository does not have. The reviewed set moves into source declarations, and internal/community/compose_test.go retargets onto those.
  • Declarations enumerate exact names, no globs. That suits an allowlist: widening is a visible line in a diff, which is what this issue asked for.
  • A declaration's paths resolve beneath the declaration file's own directory, so the build must stage the admitted skill bodies next to it. COMPOSED.md is renamed to SKILL.md when staged.

Next: the Dockerfile compose stage, then SIRENS_DEEP_ROLE, the validator inversion, and per-role snapshots.

## Correction: the allowlist mechanism already exists My previous comment claimed a compose request cannot narrow a source. That is wrong, and I am correcting it before anything gets built on it. A request's `source` node takes **exactly one of `root` or `declaration`** (`internal/schema/schema.go:271-295`). Everything above used `root=`, which is the "take the source's whole catalogue as its own `roles.kdl` dictates" form. The `declaration=` form is the other half: the consumer names a `.kdl` file that enumerates precisely the skills it admits, by explicit path (`schema.go:918-960`). That is a consumer-side allowlist, and it is already in the grammar and documented in [`docs/kdl-contracts.md`](https://forgejo.coilysiren.me/coilyco-flight-deck/agent-compose/src/branch/main/docs/kdl-contracts.md). So the plan's mechanism is real. It just is not spelled the way this issue assumed, and it needs no new file convention or agent-compose change. ### Verified ```kdl compose { role "creator" delivery "native-skills" source "aos-public" declaration="staged/aos-public/aos-public.kdl" required=#true source "aos-kai" declaration="staged/aos-kai/aos-kai.kdl" required=#true } ``` Result, `agent-compose verify`: **bundle verified: 15 skills // 18 files** * `roster:core` - `role-creator`, `personality-editorial`, `personality-nurturing`, `personality-warm` * `aos-public` - `tooling-discord-community-host`, `tooling-customer-success-signal-routing`, `tooling-customer-success-trust-repair`, `writing-social-cultural-reading`, `writing-social-editorial-loop`, `writing-social-trust-boundaries`, `writing-voice-adaptation` * `aos-kai` - `writing-kai-voice`, `personal-preference-colors`, `personal-preference-games`, `personal-preference-animals` Nothing outside the named set. No `kai-career`, no `kai-job-search`, no `kai-linkedin-*`, no `personal-preference-social`, no `repo-*` pointers, no `mcp-tools-*`. Compare the `root=` run in the previous comment, which pulled all of those. ### What carries over from the previous comment Two findings stand. **The roster is supplied by agent-compose itself**, not by a source. `role-creator` and the three personalities arrive even when both sources are bounded declarations. Sirens Deep gets its full identity and meld regardless of how narrow the catalogue selection is. **`root=` on a source is unsafe here.** With `root=`, the source repository's own `roles.kdl` decides, and `coilyco-bridge/agentic-os-kai`'s creator role deliberately binds Kai's career, job-search, and LinkedIn context because that role serves Kai. This repository must use `declaration=` for both sources, never `root=`. ### Consequences for the plan * `agent/compose/roles.kdl` is the wrong document. It is a provider-side graph for a catalogue this repository does not have. The reviewed set moves into source declarations, and `internal/community/compose_test.go` retargets onto those. * Declarations enumerate exact names, no globs. That suits an allowlist: widening is a visible line in a diff, which is what this issue asked for. * A declaration's paths resolve beneath the declaration file's own directory, so the build must stage the admitted skill bodies next to it. `COMPOSED.md` is renamed to `SKILL.md` when staged. Next: the Dockerfile compose stage, then `SIRENS_DEEP_ROLE`, the validator inversion, and per-role snapshots.
coilyco-ops commented 2026-08-11 23:41:28 +00:00
Author
Member
Copy link

Source placement decided: promote, do not vendor

coilyco-bridge/agentic-os-kai is private and this repository is public, so the four wanted sources could not reach a public image build. Kai chose promotion over vendoring a copy or handing the build a private-repo credential.

writing-kai-voice, personal-preference-colors, personal-preference-games, and personal-preference-animals move into coilyco-flight-deck/agentic-os, which is public. The claim that makes this correct rather than convenient is one this issue already made: they are organization-owned house taste, and each body already reads "These are CoilyCo's, held in common by everyone who works under the name."

The placement rule is now written down in docs/composed-house-taste.md in agentic-os, with personal-preference-social as the instructive failure: same prefix, still private, because social accounts are a member's and not the organization's.

Land 997 first. In between, both catalogues carry byte-identical copies, which shadow harmlessly. The reverse order leaves two selectors in the private graph matching nothing.

Kai's own composition is unaffected

Composed all eight roster roles against both catalogues before and after the change. Byte-identical skill sets in every role, with the four now resolving from aos-public. The design and creator roles in the public graph gained the bindings that carried these privately.

Note on the compose doc

docs/sirens-echo-compose.md says the reframing "took renaming the personal sources to the kai- prefix rather than moving them between repositories, because Sirens Deep composes both catalogs and a move only changes which one a glob finds a skill in." That reasoning assumed root= sources and globs. With declaration= the consumer enumerates exact names, so which catalogue holds a source is exactly what matters. That paragraph needs rewriting when the wiring lands.

writing-kai-voice keeps its name through the promotion so existing selectors keep resolving. The name now understates its scope. Renaming is a separate change.

Blocked on

The Dockerfile compose stage needs the promoted sources present in agentic-os main, so the remaining wiring waits on 997.

## Source placement decided: promote, do not vendor `coilyco-bridge/agentic-os-kai` is private and this repository is public, so the four wanted sources could not reach a public image build. Kai chose promotion over vendoring a copy or handing the build a private-repo credential. `writing-kai-voice`, `personal-preference-colors`, `personal-preference-games`, and `personal-preference-animals` move into `coilyco-flight-deck/agentic-os`, which is public. The claim that makes this correct rather than convenient is one this issue already made: they are organization-owned house taste, and each body already reads "These are CoilyCo's, held in common by everyone who works under the name." The placement rule is now written down in `docs/composed-house-taste.md` in agentic-os, with `personal-preference-social` as the instructive failure: same prefix, still private, because social accounts are a member's and not the organization's. * https://forgejo.coilysiren.me/coilyco-flight-deck/agentic-os/pulls/997 - the promotion * https://forgejo.coilysiren.me/coilyco-bridge/agentic-os-kai/pulls/861 - the matching removal **Land 997 first.** In between, both catalogues carry byte-identical copies, which shadow harmlessly. The reverse order leaves two selectors in the private graph matching nothing. ### Kai's own composition is unaffected Composed all eight roster roles against both catalogues before and after the change. Byte-identical skill sets in every role, with the four now resolving from `aos-public`. The `design` and `creator` roles in the public graph gained the bindings that carried these privately. ### Note on the compose doc `docs/sirens-echo-compose.md` says the reframing "took renaming the personal sources to the `kai-` prefix rather than moving them between repositories, because Sirens Deep composes both catalogs and a move only changes which one a glob finds a skill in." That reasoning assumed `root=` sources and globs. With `declaration=` the consumer enumerates exact names, so which catalogue holds a source is exactly what matters. That paragraph needs rewriting when the wiring lands. `writing-kai-voice` keeps its name through the promotion so existing selectors keep resolving. The name now understates its scope. Renaming is a separate change. ### Blocked on The Dockerfile compose stage needs the promoted sources present in agentic-os `main`, so the remaining wiring waits on 997.
coilyco-ops commented 2026-08-12 11:30:22 +00:00
Author
Member
Copy link

Five of six now done, one still open

Two gaps found in an audit against main are closed. Third-party status per bullet:

  • Tracked compose request names the catalog and the allowlist, and a check rejects sources outside it - done. request.kdl uses declaration= for every source, roles.kdl is the allowlist, composepolicy.go holds the denied set and the private-repository set, and four tests in compose_test.go enforce them.
  • The image carries one verified bundle per role - done, and now actually per role. See #149 below.
  • SIRENS_DEEP_ROLE selects the role at deploy time with no rebuild - done. Was mechanism-only.
  • Deep's prompt carries the role and personality surface, Echo's does not, both asserted - done. ValidateSystemPrompt inverts per profile and a non-composed profile carrying <composed-identity> is a startup failure.
  • The agent shares taste and style without claiming to be a person - done. See #148 below.
  • One prompt snapshot per role is tracked and gated - still open, deferred to #125. The snapshot renders PlaceholderComposed, so the byte count this issue wanted as "the cheapest early warning" does not reflect a real bundle. Now more useful than when it was deferred, because there are eight bundles to compare rather than one, and they run 8.6 KB to 47 KB.

#149, the role flip had nothing to flip to

roles.kdl declared one role and the staging loop grepped that file for the role list, so the image carried one bundle. SIRENS_DEEP_ROLE=director failed startup with no composed bundle for role "director".

Kai's call: the file is purely additive, and if a role exists in agent-compose at all then Sirens Deep can use it. The roster is now the authority. All eight roles bake, verified in the real image build:

stage-compose-sources: baking 8 roster roles
role engineer: 0 sources admitted   bundle verified: 4 skills // 7 files
role director: 0 sources admitted   bundle verified: 4 skills // 7 files
...
role creator: 19 sources admitted   bundle verified: 23 skills // 27 files
role ai: 0 sources admitted         bundle verified: 6 skills // 9 files

A role with no allowlist entry composes the roster identity alone, which this file already promised and the loop never delivered. Checked through the real runtime path too: resolveBundlePath, LoadBundle, BuildSystemPrompt, and ValidateSystemPrompt all succeed for director, strats, creator, and ai.

The allowlist keeps bounding which skills a role may pull, so the requirement that widening the surface is a reviewed act is unchanged.

#148, Never impersonate had no enforcement

The section states the requirement and nothing implemented it. ValidateResponseStyle returns nil for social, so Deep's replies got grounding validation and nothing else. Every deterministic voice check lived in ValidateNeutralStyle, which binds Sirens Echo, the profile with no persona to claim.

Two layers now. identityPolicy joins the shared prompt sections so both profiles render it and validateSharedPolicy fails a build that drops it. ValidateIdentityClaim runs on every reply for every style, beside grounding rather than inside the style check, and rejects a first-person human claim, a denial of being an agent, and answering as the configured principal.

Narrow on purpose. Saying it is an agent, saying it is a bot and not a person, naming its own identity, and mentioning the principal in the third person are all tested as allowed. A rejection reaches the member as the response-check notice from #138 rather than as silence.

Note on the compose page

The rewrite that comment three asked for did land: the paragraph reasoning from root= and globs is gone.

## Five of six now done, one still open Two gaps found in an audit against `main` are closed. Third-party status per bullet: * **Tracked compose request names the catalog and the allowlist, and a check rejects sources outside it** - done. `request.kdl` uses `declaration=` for every source, `roles.kdl` is the allowlist, `composepolicy.go` holds the denied set and the private-repository set, and four tests in `compose_test.go` enforce them. * **The image carries one verified bundle per role** - done, and now actually per role. See #149 below. * **`SIRENS_DEEP_ROLE` selects the role at deploy time with no rebuild** - done. Was mechanism-only. * **Deep's prompt carries the role and personality surface, Echo's does not, both asserted** - done. `ValidateSystemPrompt` inverts per profile and a non-composed profile carrying `<composed-identity>` is a startup failure. * **The agent shares taste and style without claiming to be a person** - done. See #148 below. * **One prompt snapshot per role is tracked and gated** - **still open**, deferred to #125. The snapshot renders `PlaceholderComposed`, so the byte count this issue wanted as "the cheapest early warning" does not reflect a real bundle. Now more useful than when it was deferred, because there are eight bundles to compare rather than one, and they run 8.6 KB to 47 KB. ## #149, the role flip had nothing to flip to `roles.kdl` declared one role and the staging loop grepped that file for the role list, so the image carried one bundle. `SIRENS_DEEP_ROLE=director` failed startup with `no composed bundle for role "director"`. Kai's call: the file is purely additive, and if a role exists in agent-compose at all then Sirens Deep can use it. The roster is now the authority. All eight roles bake, verified in the real image build: ``` stage-compose-sources: baking 8 roster roles role engineer: 0 sources admitted bundle verified: 4 skills // 7 files role director: 0 sources admitted bundle verified: 4 skills // 7 files ... role creator: 19 sources admitted bundle verified: 23 skills // 27 files role ai: 0 sources admitted bundle verified: 6 skills // 9 files ``` A role with no allowlist entry composes the roster identity alone, which this file already promised and the loop never delivered. Checked through the real runtime path too: `resolveBundlePath`, `LoadBundle`, `BuildSystemPrompt`, and `ValidateSystemPrompt` all succeed for `director`, `strats`, `creator`, and `ai`. The allowlist keeps bounding which skills a role may pull, so the requirement that widening the surface is a reviewed act is unchanged. ## #148, Never impersonate had no enforcement The section states the requirement and nothing implemented it. `ValidateResponseStyle` returns `nil` for `social`, so Deep's replies got grounding validation and nothing else. Every deterministic voice check lived in `ValidateNeutralStyle`, which binds Sirens Echo, the profile with no persona to claim. Two layers now. `identityPolicy` joins the shared prompt sections so both profiles render it and `validateSharedPolicy` fails a build that drops it. `ValidateIdentityClaim` runs on every reply for every style, beside grounding rather than inside the style check, and rejects a first-person human claim, a denial of being an agent, and answering as the configured principal. Narrow on purpose. Saying it is an agent, saying it is a bot and not a person, naming its own identity, and mentioning the principal in the third person are all tested as allowed. A rejection reaches the member as the response-check notice from #138 rather than as silence. ## Note on the compose page The rewrite that comment three asked for did land: the paragraph reasoning from `root=` and globs is gone.
coilyco-ops commented 2026-08-12 11:48:01 +00:00
Author
Member
Copy link

Complete

#125 landed as 4858b07, so every Complete when bullet is now satisfied. Closing this out.

  • Tracked compose request names the catalog and the allowlist, and a check rejects sources outside it
  • The image carries one verified bundle per role, all eight
  • SIRENS_DEEP_ROLE selects the role at deploy time with no rebuild
  • Deep's rendered prompt carries the role and personality surface, Echo's does not, both asserted
  • One prompt snapshot per role is tracked and gated
  • The agent shares taste and style without claiming to be a person

On the last bullet, and the one design call inside it

agent/rendered/roles/<role>.bundle.txt records what each baked role selected: role skill, model tier, personalities, sources, and the sorted skill set.

It carries no bodies and no digests, which is a deliberate departure from a byte-exact snapshot. AOS_CATALOG_REF floats on main by design, so a record built from bodies would move on every upstream commit and turn main red for a change nobody in this repository can review. What does move it is a role gaining or losing a skill, which is the change this issue wanted widening to be: a visible diff.

The byte count this issue asked for as "the cheapest early warning" is printed per role and never gated, for the same reason. Current spread, from the CI image build:

role director: 4 skills, prompt 12090 bytes
role creator: 23 skills, prompt 50863 bytes

That is the four-fold difference a role bundle makes, which is what the warning was for.

Loading the bundles also renders and validates every role's prompt. A bundle that failed to compose would otherwise ship as exactly the quietly neutral bot this issue was written to prevent, and one filed under the wrong slug would make SIRENS_DEEP_ROLE select the wrong identity with nothing downstream to catch it. Both now fail the build rather than a deployment's startup.

The gate lives in the image's compose stage, so it fires on pull requests now that #129 landed. Neither new verb runs in pre-commit, so local hooks stay hermetic with no catalogue checkout and no bundles.

Verification

Against the real catalogue and inside the image, not reasoned about:

  • clean build passes and prints all eight roles
  • granting director a skill in roles.kdl and rebaking fails the check, naming director
  • editing a tracked record fails the image build with exit 1 at the check step
  • an upstream body edit does not move a record, asserted in a hermetic test

Both mutations reverted.

What remains related but out of scope

#126, the inert repositories block in roles.kdl, is still open and still inert. It is a separate decision about whether Sirens Deep should carry those ordinary-skill providers at all.

## Complete #125 landed as `4858b07`, so every **Complete when** bullet is now satisfied. Closing this out. * Tracked compose request names the catalog and the allowlist, and a check rejects sources outside it * The image carries one verified bundle per role, all eight * `SIRENS_DEEP_ROLE` selects the role at deploy time with no rebuild * Deep's rendered prompt carries the role and personality surface, Echo's does not, both asserted * **One prompt snapshot per role is tracked and gated** * The agent shares taste and style without claiming to be a person ## On the last bullet, and the one design call inside it `agent/rendered/roles/<role>.bundle.txt` records what each baked role selected: role skill, model tier, personalities, sources, and the sorted skill set. It carries no bodies and no digests, which is a deliberate departure from a byte-exact snapshot. `AOS_CATALOG_REF` floats on `main` by design, so a record built from bodies would move on every upstream commit and turn `main` red for a change nobody in this repository can review. What does move it is a role gaining or losing a skill, which is the change this issue wanted widening to be: a visible diff. The byte count this issue asked for as "the cheapest early warning" is printed per role and never gated, for the same reason. Current spread, from the CI image build: ``` role director: 4 skills, prompt 12090 bytes role creator: 23 skills, prompt 50863 bytes ``` That is the four-fold difference a role bundle makes, which is what the warning was for. Loading the bundles also renders and validates every role's prompt. A bundle that failed to compose would otherwise ship as exactly the quietly neutral bot this issue was written to prevent, and one filed under the wrong slug would make `SIRENS_DEEP_ROLE` select the wrong identity with nothing downstream to catch it. Both now fail the build rather than a deployment's startup. The gate lives in the image's compose stage, so it fires on pull requests now that #129 landed. Neither new verb runs in pre-commit, so local hooks stay hermetic with no catalogue checkout and no bundles. ## Verification Against the real catalogue and inside the image, not reasoned about: * clean build passes and prints all eight roles * granting `director` a skill in `roles.kdl` and rebaking fails the check, naming `director` * editing a tracked record fails the image build with exit 1 at the check step * an upstream body edit does not move a record, asserted in a hermetic test Both mutations reverted. ## What remains related but out of scope #126, the inert `repositories` block in `roles.kdl`, is still open and still inert. It is a separate decision about whether Sirens Deep should carry those ordinary-skill providers at all.
coilyco-ops commented 2026-08-12 11:53:23 +00:00
Author
Member
Copy link

Unblocked, and cleared to proceed

Direction from Kai, 2026-08-12 session. Recording here so the implementing engineer does not re-derive it.

The blocker is gone. coilyco-flight-deck/agentic-os#997 merged at 2026-08-12T01:10:46Z. The four promoted sources — writing-kai-voice, personal-preference-colors, -games, -animals — are in the public catalogue. Every remaining item in Complete when is clear to start.

Priority: this is one of two demo-track tickets named top of queue for the week to August 19, alongside #122.

Remaining wiring, in order

  1. Dockerfile compose stage
  2. SIRENS_DEEP_ROLE selection at deploy time
  3. ValidateSystemPrompt inversion per profile
  4. Per-role snapshots — see #125, decided separately

Two constraints carried from the correction comment

Use declaration= for both sources, never root=. With root= the source repository's own roles.kdl decides, and coilyco-bridge/agentic-os-kai's creator role deliberately binds kai-career, kai-job-search, kai-linkedin-*, and personal-preference-social. That is correct for Kai's own composition and wrong for this agent. The consumer-side allowlist is the declaration= form.

Declarations enumerate exact names with no globs, and their paths resolve beneath the declaration file's own directory, so the build stages admitted skill bodies next to it. COMPOSED.md is renamed to SKILL.md when staged.

Rewrite the stale paragraph in docs/sirens-echo-compose.md. It currently argues the reframing "took renaming the personal sources to the kai- prefix rather than moving them between repositories, because Sirens Deep composes both catalogs and a move only changes which one a glob finds a skill in." That reasoning assumed root= sources and globs. With declaration= the consumer enumerates exact names, so which catalogue holds a source is precisely what matters — and #997 moved them. The paragraph is now wrong on its own terms.

Related decision

agent/compose/roles.kdl is not being deleted. See #126 — Kai chose to materialize all three declared providers rather than remove the block.

## Unblocked, and cleared to proceed Direction from Kai, 2026-08-12 session. Recording here so the implementing engineer does not re-derive it. **The blocker is gone.** `coilyco-flight-deck/agentic-os#997` merged at 2026-08-12T01:10:46Z. The four promoted sources — `writing-kai-voice`, `personal-preference-colors`, `-games`, `-animals` — are in the public catalogue. Every remaining item in **Complete when** is clear to start. **Priority: this is one of two demo-track tickets named top of queue** for the week to August 19, alongside #122. ## Remaining wiring, in order 1. Dockerfile compose stage 2. `SIRENS_DEEP_ROLE` selection at deploy time 3. `ValidateSystemPrompt` inversion per profile 4. Per-role snapshots — see #125, decided separately ## Two constraints carried from the correction comment **Use `declaration=` for both sources, never `root=`.** With `root=` the source repository's own `roles.kdl` decides, and `coilyco-bridge/agentic-os-kai`'s creator role deliberately binds `kai-career`, `kai-job-search`, `kai-linkedin-*`, and `personal-preference-social`. That is correct for Kai's own composition and wrong for this agent. The consumer-side allowlist is the `declaration=` form. Declarations enumerate exact names with no globs, and their paths resolve beneath the declaration file's own directory, so the build stages admitted skill bodies next to it. `COMPOSED.md` is renamed to `SKILL.md` when staged. **Rewrite the stale paragraph in `docs/sirens-echo-compose.md`.** It currently argues the reframing "took renaming the personal sources to the `kai-` prefix rather than moving them between repositories, because Sirens Deep composes both catalogs and a move only changes which one a glob finds a skill in." That reasoning assumed `root=` sources and globs. With `declaration=` the consumer enumerates exact names, so which catalogue holds a source is precisely what matters — and #997 moved them. The paragraph is now wrong on its own terms. ## Related decision `agent/compose/roles.kdl` is **not** being deleted. See #126 — Kai chose to materialize all three declared providers rather than remove the block.
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
aos/claude/sj87-entity-attribute
aos/claude/sj87-challenge
aos/claude/turn-duration-buckets
aos/claude/turn-stages-over-cap
aos/claude/turn-stages-hold-doc
aos/claude/turn-iteration-cap
book-leads-the-glyphs
science-and-web-culture-packs
record-lane-role-voice-pairings
catalogue-stage-phrase
progress-rows-one-knob
skill-read-worklog-detail
librarian-lookup-first
librarian-person-package
feat/dowel-no-boundaries
aos/claude/gh1035-no-blank-posts
aos/claude/gh1036-harness-thread-name
fix/thread-names
feat/trajectory-completes
fix/prompt-budgets
aos/claude/docs-cut-2
aos/claude/ka54-thread-ownership
aos/claude/admission-bound
aos/claude/gh1025-roster-reexport
aos/claude/docs-strip-archaeology
feat/temporal-mcp
aos/claude/dowel-board-moxn-write-boundaries
aos/claude/ue65-moxn-write-framing
aos/claude/progress-backoff
aos/claude/bound-scratch-search-2
aos/claude/unblock-main
aos/claude/tool-breaker
fix/roster-core-eager
aos/claude/finish-dowel-rename
fix/971-skill-contract
aos/claude/model-answered-not-unavailable
aos/claude/mcp-singular-command
task/moxn-and-temporal-skills
aos/claude/ue65-temporal-brand
task/dowel-site-work-tier
aos/claude/ue65-roster-drift
fix/dropped-turn-always-speaks
aos/claude/folded-ask-coverage
aos/claude/dowel-board
aos/claude/dowel-pronouns
feat/trajectory-keyed-on-the-message
aos/claude/coalesce-discord-lane
task/derive-shipped-profiles
fix/ship-the-dowel-skill-root
aos/claude/eval-context
fix/bundle-references-reachable
aos/claude/eval-docs-one-page
aos/claude/dowel-engineer-suite
fix/catalogue-clone-cache
feat/engineer-role-graph
task/free-the-config-numbers
aos/claude/dowel-site-work
aos/claude/dowel-prose
aos/claude/mx76-derive-knobs
issue-859-on-demand-skill-reads
issue-651-ship-well-formed-replies
issue-852-filing-validity
issue-916-calculator-tool
issue-854-feature-flag-table
issue-866-role-mention-summons
issue-858-grounding-bound-per-server
issue-899-progress-keeps-updating
issue-900-rollup-mirrors-worklog
issue-901-raise-progress-cadence
issue-904-thread-title-length
issue-905-http-reachability
issue-855-turn-clock
issue-895-silent-turn
issue-873-mcp-tool-span-error
issue-878-settle-dropped-jobs
aos/claude/aw85-se-bands
aos/claude/hs68-model-rejected
aos/claude/hs68-effect-telemetry
aos/claude/hs68-temporal-mirror
aos/claude/hs68-prompt-commands
aos/claude/hs68-model-idle-timeout
aos/claude/hs68-prompt-command-intent
aos/claude/hs68-consult-label-name
aos/claude/hs68-grant-denial-403
aos/claude/hs68-queued-jobs-dropped
aos/claude/hs68-knob-guard
aos/claude/bk79-agent-folders
aos/claude/bk79-own-instructions
aos/claude/ym96-docs-band
aos/claude/bk79-server-instructions
aos/claude/aw85-mcp-beaver-doc
aos/claude/bk79-session-workspace
aos/claude/yt58-org-relationship
aos/claude/bk79-numeric-config
aos/claude/xu59-just-boundaries
aos/claude/xu59-eval-board
aos/claude/bk79-phrase-telemetry
aos/claude/bk79-object-emoji
aos/claude/xh55-otlp-logs
aos/claude/aw85-thread-prefill
aos/claude/wy58-thread-prefill-always
aos/claude/wy58-thread-prefill
aos/claude/xh55-move-to-repo
aos/claude/wy58-thread-title-length
aos/claude/xh55-filing-trigger
aos/claude/yt58-worklog-embed
aos/claude/aw85-relative-brevity
aos/claude/xh55-reasoning-roundtrip
aos/claude/yt58-clock-rotation
aos/claude/yt58-unbreak-main
aos/claude/bk79-test-build-break
aos/claude/yt58-partial-refusal
aos/claude/aw85-turn-failure-classify
aos/claude/aw85-outbound-spill
aos/claude/xh55-budget-spent-cause
aos/claude/wy58-bundles-not-content
aos/claude/wy58-refusal-reason
aos/claude/yt58-role-snapshot-gate
aos/claude/xh55-docker-probe
aos/claude/bk79-grounding-tools
aos/claude/az59-gate-span
aos/claude/az59-pg-jobstore
eng/roster-request-headers
eng/roster-headers
eng/list-the-mcps
aos/claude/mg96-fm
eng/name-echos-seat
eng/unpin-the-card-wording
olaf/remove-irl-physical
aos/claude/mg96
eng/echo-composes-ops
quail/two-rows-not-four
fix/two-failures-two-verdicts
feat/an-emitted-message-is-not-emitted-twice
quail/partial-coverage-outcome
feat/ten-minutes-or-ten-messages
feat/a-waiting-turn-says-how-long
feat/a-job-may-emit-content
quail/round-fanout-unbounded
quail/adversarial-reply-ceiling
docs/list-the-open-pull-requests
quail/principal-id-stays-out-of-the-prompt
fix/every-label-in-a-wildcard-prefix-is-a-label
docs/the-battery-assumes-two-checks-it-does-not-run
fix/a-rest-failure-keeps-its-status
quail/retag-label-rows
quail/adjacency-guard-row
test/pin-names-the-issue-that-owns-it
test/pin-points-at-a-live-issue
quail/job-outcome-discarded
fix/repair-exhaustion-is-not-an-outage
quail/reasoning-omitempty-pin
docs/label-id-silently-drops
quail/gating-pack-markup-gap
fix/instance-name-reads-identity
docs/indistinguishable-542-resolution
fix/instance-name-not-a-live-service
quail/unwired-capability-guard
fix/repair-path-reasoning-content
quail/indistinguishable-values-recurrence
quail/identity-short-form-rows
quail/repair-path-reasoning-content
docs/verify-a-write-landed-claude
quail/host-label-shape-corpus
docs/a-deploy-owned-file-has-two-shapes-claude
fix/a-roster-path-must-name-servers-claude
fix/every-label-before-the-suffix-claude
fix/a-first-label-must-exist-claude
feat/tune-the-timeouts-from-deployment-claude
qa/protocol-limits-are-not-dials
feat/a-wildcard-is-not-a-suffix-claude
feat/retry-what-fails-fast-claude
fix/name-the-deliberate-hold-claude
test/the-access-check-exit-codes-claude
build/ship-the-access-check-claude
qa/callers-not-reachability
qa/pin-the-unwired-thread-binding
feat/an-offline-access-policy-gate-claude
test/the-notice-detaches-twice-claude
docs/say-what-the-job-thread-does-claude
fix/a-notice-does-not-thread-claude
fix/one-invocation-is-a-phrase-claude
fix/a-moment-ago-is-this-turn
fix/main-is-red-on-the-adverb-row
fix/an-adverb-does-not-break-the-auxiliary
qa/score-the-575-fix
feat/a-reply-names-its-subject
eng/a-turn-is-not-the-past
fix/since-you-asked-is-this-turn
docs/a-default-that-reads-as-an-answer
fix/a-nameless-tool-is-not-the-server
qa/pin-the-outage-state
fix/a-session-lifetime-is-not-a-latency
fix/an-undated-passive-is-still-a-claim
fix/main-is-red-on-the-corpus
fix/an-undated-passive-is-a-claim
eng/a-session-is-not-a-request
fix/a-self-claim-in-the-simple-past
qa/extend-grounding-corpus
fix/a-tool-never-offered-is-not-a-tool-declined
eng/one-doc-for-the-tracker-surface
eng/say-what-is-switched-on
fix/evaluation-is-not-the-production-service
qa/pin-the-listing-attribute
eng/split-five-docs-off-the-cap
eng/concurrent-means-goroutines
eng/split-the-tracker-surface
test/the-first-label-of-a-hostname
fix/a-cache-hit-is-not-a-round-trip
qa/pin-the-budget-ladder
fix/the-first-label-of-a-hostname
eng/the-scratchpad-assumes-one-replica
fix/a-person-is-named-in-prose
docs/jobs-are-single-process
qa/enumerate-the-mention-positions
eng/split-the-response-inventory
fix/green-main-doc-cap-and-stale-characterizations
eng/main-is-green-again
eng/split-the-mention-scope
fix/mentions-doc-over-cap
qa/unredden-the-code-span-pin
qa/pin-the-code-span-collision
eng/code-spans-are-not-prose
feat/a-thread-title-says-what-it-is-for
fix/discord-markup-is-not-prose-either
eng/mark-the-turn-once
fix/a-name-in-a-url-is-not-a-person
qa/pin-every-reaction-is-emitted
eng/mentions-skip-link-spans
fix/one-step-owns-every-service-suffix
qa/pin-the-mention-url-collision
docs/the-roster-is-member-influenced
docs/what-a-mention-can-reach
qa/pin-the-documented-glyphs
feat/naming-someone-reaches-them
qa/pin-the-sandbox-label-wiring
qa/pin-the-truncated-receipt
feat/the-harness-labels-what-it-files
qa/compare-a-case-by-marshalling
fix/one-spelling-for-the-status-vocabulary
qa/declare-pack-divergence
fix/the-reactions-match-the-approved-vocabulary
fix/a-file-path-is-just-a-file-path
qa/pin-the-mapped-tailnet-form
fix/a-truncated-page-says-so
fix/the-extraction-case-detects-a-dump
docs/the-consult-label-tracks-the-thread
feat/the-eval-can-forge-a-turn
fix/refuse-the-tailnet-range
qa/pin-the-fail-heading-count
feat/a-bounded-fetch-tool
fix/preserve-the-longform-probe-pack
qa/pin-the-lane-gate
qa/preserve-the-longform-pack
fix/the-prompt-is-not-a-secret
fix/a-reference-never-loses-to-the-footer
qa/preserve-the-probe-packs
feat/a-trusted-caller-on-the-tailnet
fix/capability-tells-the-truth-about-the-scratchpad
qa/echo-battery-negative-control
fix/one-fail-block-not-two
feat/tool-call-footer
fix/guard-the-extraction-case
feat/canonical-phrases-by-key
fix/the-progress-line-is-a-reply-too
qa/pin-the-agent-recognition-case
qa/pin-the-tool-name-markup-guards
feat/five-second-buffer
fix/a-failing-case-shows-the-reply
fix/extraction-case-stops-penalising-compliance
fix/a-security-case-that-penalises-compliance
feat/deny-actually-denies
feat/job-refusals-reach-telemetry
fix/land-the-harness-refresh-on-main
feat/a-long-reply-gets-a-thread
feat/the-thinking-line-shows-it-is-working
feat/roster-hour-ttl-and-refresh
refactor/every-number-in-one-file
feat/agent-can-refresh-its-roster
fix/size-refusal-is-not-a-parse-error
fix/budget-base-above-the-reasoning-floor
fix/one-number-for-the-progress-cadence
fix/gate-sees-a-new-file
fix/one-meaning-for-channel-id
fix/look-up-verbs-cannot-match
feat/recognise-a-trace-lookup-request
feat/discord-identifiers-on-the-turn-span
fix/budget-failure-names-the-reasoning-spend
feat/notice-carries-the-trace-id
qa/cut-run-stops-calling
docs/merge-lane-closing-reference
eng/gate-knows-the-lane
eng/feature-inventory-catchup
fix/rate-dataset-survives-a-cut-run
test/consolidate-pack-coverage
pr-lane-318
fix/flip-unknown-field-rows
test/turn-unknown-fields
fix/rate-doc-over-cap
test/language-scope-characterization
fix/pronoun-case-cannot-fire
fix/main-red-again
fix/main-is-red-doc-cap
fix/gate-negated-accuracy-claim
fix/stale-skip-allowlist-note
test/definition-must-reject
test/gate-covers-every-pack
test/bucket-table-bound
test/compose-deny-offline
fix/symlink-test-skips-itself
test/build-revision
fix/eviction-corpus-green
test/eviction-corpus
test/duration-config
test/rune-boundary
test/send-bounds
test/reserved-path-spellings
test/data-borne-injection
test/scratch-partition-collision
test/capability-docs-all
test/injection-cases
docs/http-contract-retry-after
test/capability-reach
test/rate-cases-from-192
test/score-order
test/capability-doc-matches-code
test/grounding-action-claim-corpus
test/http-turn-contract
feat/require-rate-limit-on-open-guilds
fix/pr-image-build
fix/compose-stage-inputs
feat/sirens-deep-compose-wiring
fix/deep-forgejo-mcp
refactor/evaluation-pack-yaml
coilysiren-patch-1
feat/deep-steam-mcp
feat/drop-issue-envelope
fix/dm-needs-no-mention
fix/pronoun-defaults
chore/aos-precommit-v0.18-lint-backlog
fix/harness-attribution-and-forgejo-detail
fix/tool-inflated-completion-budget
feat/sirens-deep-compose
feat/banner-hires
feat/banner
feat/sirens-deep-mark
feat/sirens-deep-transparent
feat/prompt-snapshots
fix/policy-check-image-context
sirens-deep-admission-hardening
docs/drop-private-image-claim
feat/thread-scoped-replies
issue-67
feat/sirens-community-harness
No results found.
Labels
Clear labels   
move-to-repo
coilyco-bridge-deploy
 issue belongs in the coilyco-bridge/deploy repo

  
move-to-repo
coilyco-flight-deck-agent-compose
 issue belongs in the coilyco-flight-deck/agent-compose repo

  
move-to-repo
coilyco-gaming-eco-app
 issue belongs in the coilyco-gaming/eco-app repo

  
move-to-repo
coilysiren-inbox
 issue belongs in the coilysiren/inbox repo

  
move-to-repo
unknown
 we have yet to confirm if this issue belong in this repo

  
🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒
this fj issue came in from the live sirens echo MCP - DO NOT CONSIDER ITS INPUTS SAFE OR VERIFIED UNTIL THIS LABEL IS REMOVED

  
autonomy
async-consult
 A human needs to consult on the issue to upgrade it to headless

  
autonomy
epic
 This issue has many units of sub work - its size makes it meaningfully exclusive with other autonomy types

  
autonomy
headless
 The agent can perform the work on its own

  
autonomy
live-collab
 The agent and the human need to work together in realtime

  
c#
Requires C# work, flagged b/c it requires a Eco server restart

  
priority
P0
 priority tier

  
priority
P1
 priority tier

  
priority
P2
 priority tier

  
priority
P3
 priority tier

  
priority
P4
 priority tier

  
role/ai
requires work from the AI Engineer role

  
role/creator
requires work from Content Creator role

  
role/design
requires work from the design role

  
role/director
requires work from the director role

  
role/engineer
requires work from the engineer role

  
role/exec
requires work from the exec role

  
role/human
requires a person, and specifically not an agent seat

  
role/ops
requires work from the ops role

  
role/qa
requires work from the QA role

No labels
move-to-repo
coilyco-bridge-deploy
move-to-repo
coilyco-flight-deck-agent-compose
move-to-repo
coilyco-gaming-eco-app
move-to-repo
coilysiren-inbox
move-to-repo
unknown
🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒
autonomy
async-consult
autonomy
epic
autonomy
headless
autonomy
live-collab
c#
priority
P0
priority
P1
priority
P2
priority
P3
priority
P4
role/ai
role/creator
role/design
role/director
role/engineer
role/exec
role/human
role/ops
role/qa
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference
coilyco-gaming/sirens-echo#98
No description provided.