Promote k3s triage trees and trap docs into Ops investigation skills #704

Closed
opened 2026-07-31 08:54:01 +00:00 by coilyco-ops · 0 comments
Owner

Outcome

Turn the symptom-driven k3s deployment triage material into discoverable Ops investigation skills while retaining architecture and configuration explanation as durable docs.

Depends on the canonical provider layout in #702

Source material

Audit and consolidate the operator procedures currently spread across:

  • docs/k3s-deploy-notes-triage-tree.md
  • docs/k3s-deploy-notes-triage-tree-runner.md
  • docs/k3s-deploy-notes-triage-tree-tailscale-proxy.md
  • docs/k3s-deploy-notes-traps.md
  • docs/k3s-deploy-notes-traps-host.md
  • docs/k3s-deploy-notes-traps-caddy.md
  • docs/k3s-deploy-notes-traps-runner.md
  • docs/k3s-deploy-notes-traps-forgejo.md
  • docs/k3s-deploy-notes-traps-publish-image.md
  • docs/k3s-deploy-notes-traps-tailscale-proxy.md
  • docs/k3s-deploy-notes-traps-terraform.md

Skill shape

  • Create the smallest useful router and focused investigation skills for the actual symptom families. Do not create one skill per docs file.
  • Cover workload startup failures, image pulls, ExternalSecret or certificate failures, host and Caddy routing, Forgejo runner failures, publish-image failures, and Tailscale proxy failures.
  • Reuse the existing pod-eviction and k3s-upgrade skills rather than copying their procedures.
  • Keep descriptions focused on fuzzy symptom discovery and aliases. Move detailed branching into references.
  • Preserve the cluster ownership boundary. Infrastructure skills diagnose cluster and host behavior. Deploy skills remain authoritative for service rollout configuration.
  • Separate approved read-only observation from mutations, rollout, cleanup, and secret handling.
  • Name the evidence and stop condition for every branch before recommending a change.
  • Keep docs/k3s-deploy-notes.md as the top-level architecture and deployment reference, with operator branches routed into skills.
  • Remove duplicated procedures and repair all navigation links.

Acceptance

  • Common prompts such as pod crashloop, image pull failure, ExternalSecret not syncing, runner job stuck, publish image failure, Caddy route broken, and Tailscale proxy unreachable each discover a clear skill.
  • Each symptom has one authoritative investigation branch.
  • Existing pod-eviction and k3s-upgrade skills remain peers with explicit routing rather than duplicated content.
  • Skill references preserve the known traps and concrete evidence checks from the source docs.
  • The skills do not grant live mutation authority or expose secret values.
  • Architecture, topology, and configuration-contract material that remains in docs is not copied into skill bodies.
  • docs/FEATURES.md reflects any materially reshaped operator capability.
  • Skill validation, dead-link checks, and pre-commit run --all-files pass.
## Outcome Turn the symptom-driven k3s deployment triage material into discoverable Ops investigation skills while retaining architecture and configuration explanation as durable docs. Depends on the canonical provider layout in https://forgejo.coilysiren.me/coilyco-flight-deck/infrastructure/issues/702 ## Source material Audit and consolidate the operator procedures currently spread across: * `docs/k3s-deploy-notes-triage-tree.md` * `docs/k3s-deploy-notes-triage-tree-runner.md` * `docs/k3s-deploy-notes-triage-tree-tailscale-proxy.md` * `docs/k3s-deploy-notes-traps.md` * `docs/k3s-deploy-notes-traps-host.md` * `docs/k3s-deploy-notes-traps-caddy.md` * `docs/k3s-deploy-notes-traps-runner.md` * `docs/k3s-deploy-notes-traps-forgejo.md` * `docs/k3s-deploy-notes-traps-publish-image.md` * `docs/k3s-deploy-notes-traps-tailscale-proxy.md` * `docs/k3s-deploy-notes-traps-terraform.md` ## Skill shape * Create the smallest useful router and focused investigation skills for the actual symptom families. Do not create one skill per docs file. * Cover workload startup failures, image pulls, ExternalSecret or certificate failures, host and Caddy routing, Forgejo runner failures, publish-image failures, and Tailscale proxy failures. * Reuse the existing pod-eviction and k3s-upgrade skills rather than copying their procedures. * Keep descriptions focused on fuzzy symptom discovery and aliases. Move detailed branching into references. * Preserve the cluster ownership boundary. Infrastructure skills diagnose cluster and host behavior. Deploy skills remain authoritative for service rollout configuration. * Separate approved read-only observation from mutations, rollout, cleanup, and secret handling. * Name the evidence and stop condition for every branch before recommending a change. * Keep `docs/k3s-deploy-notes.md` as the top-level architecture and deployment reference, with operator branches routed into skills. * Remove duplicated procedures and repair all navigation links. ## Acceptance * Common prompts such as pod crashloop, image pull failure, ExternalSecret not syncing, runner job stuck, publish image failure, Caddy route broken, and Tailscale proxy unreachable each discover a clear skill. * Each symptom has one authoritative investigation branch. * Existing pod-eviction and k3s-upgrade skills remain peers with explicit routing rather than duplicated content. * Skill references preserve the known traps and concrete evidence checks from the source docs. * The skills do not grant live mutation authority or expose secret values. * Architecture, topology, and configuration-contract material that remains in docs is not copied into skill bodies. * docs/FEATURES.md reflects any materially reshaped operator capability. * Skill validation, dead-link checks, and pre-commit run --all-files pass.
Sign in to join this conversation.
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/infrastructure#704
No description provided.