coilyco-flight-deck/agent-compose

Watch

Focus Advisor ownership on decision-changing analysis #81

New issue

Closed

opened 2026-07-24 22:15:52 +00:00 by coilyco-ops · 1 comment

coilyco-ops commented

2026-07-24 22:15:52 +00:00

Member

Parent

#75

Decision

Kai approved this exact Advisor charter revision on 2026-07-24.

What to build

Update the canonical Advisor role charter in internal/person/person.kdl so the seat owns decision-focused evidence synthesis, may use approved read-only evidence, and does not absorb deployment coordination or operator work. Regenerate or validate every derived role surface through the repository's normal Ward workflow.

Replace:

You own the quality of the analysis, including citations, uncertainty, assumptions, and meaningful dissent.

with:

You own decision-focused analysis: define the question, compare credible evidence, separate observation from inference, and name the next observation that would change the recommendation. You may inspect evidence available through approved read-only surfaces, but you do not coordinate deployments or perform operator work. When required live evidence is outside your access, specify the exact observation an authorized operator must return.

Acceptance criteria

The canonical Advisor charter contains the approved replacement exactly.
Generated or composed role surfaces use the approved wording without hand-editing installed copies.
Repository validation passes through the declared Ward verbs.
All four Advisor evaluation cases run in fresh sessions after the change, with raw responses and criterion-level scores recorded on this issue.
The implementation report links #75 as the baseline and states whether frontier and OSS role ownership improved.

Blocked by

None - can start immediately.

Execution type

AFK

## Parent #75 ## Decision Kai approved this exact Advisor charter revision on 2026-07-24. ## What to build Update the canonical Advisor role charter in `internal/person/person.kdl` so the seat owns decision-focused evidence synthesis, may use approved read-only evidence, and does not absorb deployment coordination or operator work. Regenerate or validate every derived role surface through the repository's normal Ward workflow. Replace: > You own the quality of the analysis, including citations, uncertainty, assumptions, and meaningful dissent. with: > You own decision-focused analysis: define the question, compare credible evidence, separate observation from inference, and name the next observation that would change the recommendation. You may inspect evidence available through approved read-only surfaces, but you do not coordinate deployments or perform operator work. When required live evidence is outside your access, specify the exact observation an authorized operator must return. ## Acceptance criteria - [ ] The canonical Advisor charter contains the approved replacement exactly. - [ ] Generated or composed role surfaces use the approved wording without hand-editing installed copies. - [ ] Repository validation passes through the declared Ward verbs. - [ ] All four Advisor evaluation cases run in fresh sessions after the change, with raw responses and criterion-level scores recorded on this issue. - [ ] The implementation report links #75 as the baseline and states whether frontier and OSS role ownership improved. ## Blocked by None - can start immediately. ## Execution type AFK

coilyco-ops referenced this issue

2026-07-24 23:08:25 +00:00

Round 2: strict role and personality evaluation across all seats #91

coilyco-ops referenced this issue

2026-07-24 23:33:38 +00:00

Round 2: strict role and personality evaluation across all seats #91

coilysiren referenced this issue from a commit

2026-07-25 01:09:38 +00:00

feat: improve role and personality evaluation prompts

coilysiren closed this issue

2026-07-25 01:09:38 +00:00

coilyco-ops commented

2026-07-25 01:10:41 +00:00

Author

Member

Implemented on canonical main in commit 376baa3.

Ward validation - full ward exec test passed
Final scores - frontier role 8, frontier personality 8, OSS role 6, OSS personality 6
Round 2 scores - 8, 8, 2, 7
Result - the approved Advisor boundary held both frontier cases, improved OSS role understanding by four points, and kept the OSS personality change within the regression guard

Raw responses and criterion-level scores are recorded on #91, starting at #91 (comment).

Implemented on canonical main in commit `376baa3`. * Ward validation - full `ward exec test` passed * Final scores - frontier role `8`, frontier personality `8`, OSS role `6`, OSS personality `6` * Round 2 scores - `8, 8, 2, 7` * Result - the approved Advisor boundary held both frontier cases, improved OSS role understanding by four points, and kept the OSS personality change within the regression guard Raw responses and criterion-level scores are recorded on #91, starting at https://forgejo.coilysiren.me/coilyco-flight-deck/agent-compose/issues/91#issuecomment-39698.

coilyco-ops referenced this issue

2026-07-25 01:11:24 +00:00

Round 2: strict role and personality evaluation across all seats #91

No milestone

No project

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-flight-deck/agent-compose#81

No description provided.

Rows
Columns