Remove foreign skill projection from aoa-evals#427
Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: f3779693b2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 30189c8d00
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
1 participant
What changed
.agents/skills/packages whose metadata points to8Dionysus/aoa-skillsaoa-evalswithout a top-levelskills/home because no repo-specific candidate has passed the admission threshold.agents/skills/when no owner-admitted top-levelskills/home existsWhy
The repository carried copied shared skills as if they were local support surfaces. Fresh prompt inspection showed that ten of those foreign copies entered the model-visible routing catalog. This duplicated shared delivery, created routing competition, and did not establish an eval-owned capability.
The consolidated skill model keeps shared procedure truth in
aoa-skillsand admits a repo-owned skill only after manual trigger, ABI, composition, coexistence, and no-skill evidence.Impact
Fresh
aoa-evalssessions no longer advertise foreign repo-local skills. Current user/system skills remain visible. No eval bundle, verdict, proof claim, or runtime service behavior changes.Validation
codex debug prompt-inputinspection: zero repo-local skills after removal; current user/system skills remain visible.agents/skills/withoutskills/is rejectedpython scripts/generate_decision_indexes.py --checkpython scripts/validate_semantic_agents.pypython scripts/validate_nested_agents.pypython scripts/validate_repo.pypython -m pytest -q tests/test_root_surface_roles.py tests/test_script_topology.py tests/test_validation_topology.py(78 passed,783 subtests passed)python -m pytest -q(915 passed,3 skipped,1767 subtests passed)