AAgentFile Check

REPOSITORY INSTRUCTIONS · STATIC REVIEW

Your AGENTS.md should be operational, not inspirational.

Paste repository instructions and get an explainable quality report: commands, scope, testing, safety, contradictions, context cost, and concrete language.

No uploadActionabilityConflict checksToken estimate

INSTRUCTION AUDIT

AGENTS.md linter

AGENTS.md0 words · ~0 tokens

Checks structure, commands, testing, boundaries, ambiguity, conflicts, secret patterns, and instruction density.

01

Can an agent act?

Finds concrete install, build, lint, test, and verification commands instead of generic advice.

02

Where does it apply?

Looks for repository scope, nested-file precedence, generated files, and protected paths.

03

Can rules conflict?

Flags unconditional absolutes and opposing directions that make compliance impossible.

04

Is it safe to share?

Detects credential-like strings and risky destructive or bypass commands inside examples.

QUALITY MODEL

Good context reduces agent guesswork.

AGENTS.md is intentionally plain Markdown. That means quality cannot be reduced to schema validation. The useful question is whether an unfamiliar coding agent can identify the workspace, make a scoped change, run the right checks, and know when it is done.

High-value content

  • Exact setup and verification commands
  • Repository layout and ownership boundaries
  • Generated-file and migration rules
  • Test expectations proportional to changes
  • PR and handoff requirements

Common smells

  • “Follow best practices” without examples
  • Huge pasted architecture documentation
  • Contradictory always/never rules
  • Commands that bypass safety checks
  • Tool-specific rules presented as universal
AGENTS.md generatorBuild a concise starting file →AGENTS.md best practicesWrite instructions agents can execute →CLAUDE.md vs AGENTS.mdChoose scope and precedence →