CLI: make AGENTS-relative skill paths unambiguous when cwd is a nested checkout
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.4k
- PR merge metrics
- PR metrics pending
Description
Summary
When an active AGENTS.md refers to a project-installed skill as .agents/skills/<name>/SKILL.md, an agent running from a nested checkout can incorrectly resolve that path against its process cwd instead of the directory that contains the applicable AGENTS.md.
Reproduction
- Arrange a workspace as follows:
workspace/
├── AGENTS.md
├── .agents/skills/beads/SKILL.md
└── checkout/
- Start Codex with
workspace/checkoutas the cwd while the parentworkspace/AGENTS.mdis active. - Have
AGENTS.mddirect the agent to load.agents/skills/beads/SKILL.md.
Actual behavior
The relative path is easy to interpret as workspace/checkout/.agents/..., which does not exist, even though the skill registry exposes the canonical absolute skill location.
Expected behavior
Codex should expose an unambiguous canonical origin for every active AGENTS.md and canonical absolute paths for registered skills, or document/enforce that paths referenced by AGENTS.md are resolved relative to that file rather than cwd. This prevents unnecessary failed filesystem probes and makes project-level skill discovery deterministic.
Suggested direction
A codex debug agents-md/skill-resolution view could report the instruction-file origin and the resolved path for each referenced project skill.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Start by tracing AGENTS.md discovery and project-skill resolution from the CLI, then reproduce the nested-checkout layout with workspace/checkout as cwd. Done means the instruction-file origin and registered skill paths are unambiguous, either through canonical output such as a debug view or documented enforcement of resolution relative to AGENTS.md.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- rust
- Domain
- cli
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Active
- Clarity
- Mostly clear
- Newbie friendliness
- 52/100