Skip to content

Add repo-contained agent skills#48

Open
Anemll wants to merge 1 commit into
mainfrom
codex/repo-skills
Open

Add repo-contained agent skills#48
Anemll wants to merge 1 commit into
mainfrom
codex/repo-skills

Conversation

@Anemll

@Anemll Anemll commented Mar 8, 2026

Copy link
Copy Markdown
Owner

Summary

  • move Codex skill references to repo-contained files under agents/skills/
  • add repo-local divergence and lowering skill docs for released workflows on main
  • add a public tests/dev/DIVERGENCE_HARNESS.md guide and update Claude guidance to reference repo-local agent resources

Notes

  • excludes macOS UI automation guidance from this repo
  • excludes unreleased AQ1-specific workflow references

Testing

  • not run (documentation and agent workflow scaffolding only)

Note

Low Risk
Low risk: changes are documentation/agent workflow scaffolding only, with no code-path, model conversion, or runtime behavior modifications.

Overview
Moves agent skill references from machine-local paths to repo-contained agents/skills/ docs so different agents can share the same workflows.

Adds two new skill playbooks—ane-divergence-analysis and ane-lowering-plan—and introduces tests/dev/DIVERGENCE_HARNESS.md as a public guide for running the released divergence/instability harnesses. Updates AGENTS.md and CLAUDE.md to point at these repo-local resources and drops the macOS UI automation skill reference.

Written by Cursor Bugbot for commit 7e07b76. This will update automatically on new commits. Configure here.

@cursor cursor Bot left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Cursor Bugbot has reviewed your changes and found 1 potential issue.

Bugbot Free Tier Details

You are on the Bugbot Free tier. On this plan, Bugbot will review limited PRs each billing cycle.

To receive Bugbot reviews on all of your PRs, visit the Cursor dashboard to activate Pro and start your 14-day free trial.

Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.

- `tests/dev/gemma3_divergence_harness.py`
- `tests/dev/test_gemma3_coreml_chunks_vs_pytorch.py`
- `docs/GEMMA3_FP16_SCALING.md`
- `anemll/utils/fp16_compatibility_check.py`

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Lowering plan duplicates entire divergence analysis workflow

Low Severity

The ane-lowering-plan/SKILL.md embeds a near-complete copy of the divergence analysis workflow (decision tree, preflight checklist, core commands, metrics interpretation, triage steps, and probe checks) that already has its own dedicated skill file in ane-divergence-analysis/SKILL.md. This creates a maintenance burden — any update to the divergence workflow needs to be applied in two places, with inconsistencies likely over time. The lowering plan skill's description also claims it handles divergence analysis tasks, which conflicts with AGENTS.md trigger rules that route tasks to the matching skill, creating ambiguity about which skill an agent will select. A cross-reference to the divergence analysis skill would avoid the duplication.

Additional Locations (1)

Fix in Cursor Fix in Web

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant