Agents / scripts
AGENTS.md1 starsChanged 58 days agoApache-2.0View on GitHub
Operational tooling and CI guards. Nothing here ships; everything here can fail a merge. This directory is the repository's enforcement surface: feature proofs, drift guards, corpus generators and the merge-gate seeding pipeline. It is not part of any installed package and carries its own gate. README.md indexes every script by purpose; this file is what that index does not tell an agent about to edit one. - validations/ is protected. Changes there need the eval-change-approved label and a code owner.…
In this repository39 other files it gives its agents.
- agent-core/AGENTS.mdAGENTS.md
- .agents/AGENTS.mdAGENTS.md
- AGENTS.mdAGENTS.md
- behavioral-regression/AGENTS.mdAGENTS.md
- claude-foundation/AGENTS.mdAGENTS.md
- config/AGENTS.mdAGENTS.md
- corpora/AGENTS.mdAGENTS.md
- demo/AGENTS.mdAGENTS.md
- docs/AGENTS.mdAGENTS.md
- examples/AGENTS.mdAGENTS.md
- experiments/AGENTS.mdAGENTS.md
- flow-corpus/AGENTS.mdAGENTS.md
- flow-protocol/AGENTS.mdAGENTS.md
- openspec/AGENTS.mdAGENTS.md
- skills/AGENTS.mdAGENTS.md
- tests/AGENTS.mdAGENTS.md
- claude-foundation/CLAUDE.mdCLAUDE.md
- merge-gate-auditorSkill
- update-executive-reportSkill
- update-executive-reportSkill
- architecture-drift-guardSkill
- commonSkill
- corpus-guardianSkill
- dataset-lintSkill
- deploySkill
- e2e-matrix-sentinelSkill
- eval-corpus-forgeSkill
- hierarchical-recursive-brainstormSkill
- model-benchSkill
- openai-judgeSkill
- openspec-implementation-reviewSkill
- openspec-peer-reviewSkill
- openspec-quality-planSkill
- pre-pr-gateSkill
- project-setupSkill
- quality-gateSkill
- refactoring-decomposerSkill
- repo-invariant-reviewSkill
- test-completeness-guardSkill
The file
Files
AGENTS.md — scripts
Operational tooling and CI guards. Nothing here ships; everything here can fail a merge.
This directory is the repository's enforcement surface: feature proofs, drift guards, corpus
generators and the merge-gate seeding pipeline. It is not part of any installed package and
carries its own gate. README.md indexes every script by purpose; this file is
what that index does not tell an agent about to edit one.
Map
| Path | Role |
validate.py | Drives features.yaml; runs the tier-filtered F_0NN.py proofs |
validations/ | 69 one-shot feature proofs. PROTECTED; a gate, not an editable surface |
quality-gate.sh | GENERATED. Hand extensions only in do_extra() below the marker |
validate_skill.py | Canonical copy; vendored byte-identically into every skill |
eval_protected_paths.py | Single source of truth for the protected-path globs |
.coveragerc | This directory's own coverage floor of 85 |
_cli.py, _config.py, _provenance.py, _*_corpus_lib.py | Shared libraries; no entry points |
Diagram
flowchart LR
accTitle: scripts ownership and generated artifacts
accDescr: features.yaml drives validate.py into the protected one-shot proofs, the quality-gate skill generates quality-gate.sh, and the canonical validate_skill.py is copied into every skill under a drift guard.
FEAT["features.yaml"]
GEN["skills/quality-gate/<br/>gen_gate.py"]
VEND["each skill's scripts/<br/>vendored copies"]
subgraph SC["scripts/ (you are here)"]
VAL["validate.py"]
PROOF["validations/F_0NN.py<br/>PROTECTED"]
GATE["quality-gate.sh<br/>GENERATED"]
CANON["validate_skill.py"]
DRIFT["check_skill_script_drift.py"]
RC[".coveragerc<br/>floor 85"]
end
FEAT --> VAL
VAL --> PROOF
GEN --> GATE
GATE --> RC
CANON --> VEND
DRIFT --> CANON
DRIFT --> VEND
classDef here fill:#e8f0fe,stroke:#1a73e8,stroke-width:2px
class SC hereRules that bite here
validations/is protected. Changes there need theeval-change-approvedlabel and a code owner. The files are one-shot proofs a feature stayed true, invoked byvalidate.pyfromfeatures.yaml— not unit tests, which is why.coveragercomits them from the floor.quality-gate.shis generated; never hand-edit above the marker. Regenerate withpython skills/quality-gate/scripts/gen_gate.py --root . --typecheck-path src/eval_harness --typecheck-path scripts --typecheck-path tests. Onlydo_extra()below the marker survives a regeneration, and that is where the F-031 scripts-coverage stage lives.validate_skill.pyexists in many copies. Editing the canonical one without copying it into everyskills/<name>/scripts/failscheck_skill_script_drift.py.- This directory has its own floor of 85, read from
.coveragercby the gate'sdo_extra()stage..coveragercis itself protected and pinned in../coverage-floors.yaml; lowering it is a reviewed act. - Invoke mypy as
python3 -m mypy, never a baremypyfromPATH. A stray install shadows the pinned one, runs without this repo's dependencies, and reports missing-import errors that look like real defects and vanish under the correct interpreter.
Verify
./scripts/quality-gate.sh allSubagents
| Task in this directory | Agent | Why |
| Find every call site of a guard before changing its signature | explorer | Read-only Grep sweep; these modules are imported by tests, hooks and workflows alike |
| Run the gate and isolate which stage failed | test-runner | Has Bash; the gate has four stages and only names the failing one in its own output |
| Review a guard change before pushing | narrow-critic | A guard that fails open is a security defect a linter cannot see |
See also
| Doc | Read it when |
README.md | You need to know what an unfamiliar script in this directory does |
eval_protected_paths.py | You are about to edit a file and need to know whether it needs the approval label |
../coverage-floors.yaml | A coverage floor is in your way and you are tempted to lower it |
../docs/decisions/0019-size-budget-gate.md | A script here is nearing 500 lines and you need the split rule |
../skills/AGENTS.md | Your change touches a vendored skill script or the marketplace |
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

