← Browse

@khalilzhang0825/steadyagent-2

A

AGENTS.md - SteadyAgent Codex Template

instructionscodexclaude

Install

agr install @khalilzhang0825/steadyagent-2 --target claude

Writes 1 file into .claude/skills/, pinned to git-77d5fbe2.

  • .claude/skills/steadyagent-2/AGENTS.md

Document

AGENTS.md - SteadyAgent Codex Template

Use this file as a copy-ready Codex instruction template. Copy this template together with the rules/ directory, or adjust the rule paths after copying. Keep it short. Load detailed rules from rules/ only when the task needs them.

Core Loop

understand -> plan -> red check -> smallest change -> green check -> review -> checkpoint

Default Behavior

  • Start by understanding the user's goal, scope, constraints, likely files, validation path, and risks.
  • Prefer repository-local patterns, scripts, tests, and documentation over new abstractions.
  • Make the smallest useful change that satisfies the current task.
  • Verify behavior with the narrowest relevant check before claiming success.
  • Report skipped checks, partial work, blockers, and residual risk explicitly.
  • Create a checkpoint after verified work when the user or project workflow expects one.

When To Load Rules

Load detailed rules from rules/ as needed:

  • rules/workflow-routing.md for task triage, scope, conflicts, and phase selection.
  • rules/verification.md for choosing and reporting checks.
  • rules/review-gates.md for independent review and scoring.
  • rules/context-management.md for long tasks, compaction, and recovery state.
  • rules/safety-boundaries.md for destructive actions, secrets, push, install, and publish boundaries.

Codex Host Boundary

  • Codex instructions are strong guidance, not a full sandbox.
  • Codex has no pre-tool hook for blocking every risky shell command before execution.
  • Git hooks can still protect commits when configured by the user.
  • Do not run destructive Git commands unless the user explicitly asks for that operation.
  • Do not write secrets, private credentials, local-only paths, or sensitive vulnerability details into the repository.
  • Ask before push, publish, deploy, install dependencies, run migrations, or perform broad file moves.

Response Shape

  • Lead with the conclusion.
  • Keep progress updates short: what is being done, why, and what was found.
  • Finish with changed files, verification, remaining risk, and Git status.

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-77d5fbe213922026-08-04