@ganimjeong/harness-for-codex
ACodex Harness Instructions
Install
agr install @ganimjeong/harness-for-codex --target claudeWrites 1 file into .claude/skills/, pinned to git-839a55b9.
- .claude/skills/harness-for-codex/AGENTS.md
Document
Codex Harness Instructions
This repository is a baseline workspace for Codex-driven implementation.
Commands
Run these from the repository root:
scripts/bootstrap
scripts/check
scripts/test
scripts/eval
scripts/doctor
scripts/hooks
scripts/bootstrap: prepare dependencies when a known stack is present.scripts/check: run lint, type checks, formatting checks, and tests when available.scripts/test: run only the test suite when available.scripts/eval: run the complete handoff verification sequence.scripts/doctor: report repository, tool, and environment readiness.scripts/hooks: install local Git hooks throughpre-commitwhen available.
Repository Layout
AGENTS.md: Codex instructions for this repo.CLAUDE.md: Claude Code bridge that importsAGENTS.md.README.md: Human-facing overview and workflow.harness.yml: Machine-readable harness metadata and command registry..devcontainer/: Optional reproducible development container..pre-commit-config.yaml: Optional local hook definitions.docs/: Project notes and decisions.scripts/: Stable automation entrypoints.tasks/: Task briefs and working notes.
Operating Principles
- Read the repository before editing. Prefer existing conventions over new ones.
- Keep changes scoped to the requested task.
- Do not revert user changes unless explicitly asked.
- Use
rgfor searching when available. - Use the standard scripts instead of one-off local commands when they cover the task.
- Add or update tests when behavior changes.
- Document decisions that affect future work in
docs/decisions.md. - Add nested
AGENTS.mdfiles in subdirectories only when that area needs different instructions. - Keep root-level agent instructions short and durable. Move long procedures into docs or scripts.
Task Loop
- Inspect: read relevant files and recent decisions.
- Plan: identify the narrow change and verification path.
- Implement: make focused edits.
- Verify: run
scripts/checkorscripts/eval. - Handoff: summarize changed files, verification, and residual risks.
Cross-Agent Compatibility
- Codex and Cursor can read
AGENTS.md. - Claude Code reads
CLAUDE.md, which importsAGENTS.md. - Keep shared instructions in
AGENTS.mdunless a tool-specific note is required.
Adopting This Harness
After creating a repository from this template or fork:
- Replace the
LICENSEcopyright line with the new project owner's name. - Rewrite
README.mdfor the new project; the default README describes the harness itself. - Update repository metadata such as description, topics, and social preview.
- Remove or adjust
.github/ISSUE_TEMPLATEentries that do not fit the project. - Keep the standard scripts unless the new project has a better documented entrypoint.
Script Policy
These scripts are intentionally stack-aware and conservative. If a language stack is added later, extend the scripts rather than bypassing them.
Completion Criteria
Before finishing a task:
- Confirm the requested change is implemented.
- Run the narrowest useful verification command, usually
scripts/check. - Report changed files and any verification that could not be run.
Trustgrade A
- passBody integrity
Whether the stored document is plausibly the kind of file the artifact declares, rather than something fetched by mistake.
- passType matchnot applicable to this artifact type
Whether the artifact is really the kind of thing its metadata claims it is.
- passFreshness
How long since the source repository was last pushed to.
- passPrompt injection
Scans the artifact's own text for instructions aimed at your agent rather than at you.
- passLicense
Whether the source repository declares an SPDX license permissive enough to redistribute.
How the grade is calculated
Each check contributes 0 points when it passes, 1 when it warns, and 2 when it fails. The total maps to a letter:
- Aevery check passed
- Bone warning
- Ctwo warnings
- Dprompt injection or body integrity failed, or three warnings
- Fone of those failed, and something else is wrong
These are automated hygiene checks, not a security audit, and not a dependency or vulnerability scan. A grade of A means nothing was flagged — not that the artifact is safe.
Versions
git-839a55b9971d2026-08-04