subagent-creator — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited subagent-creator (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Create Codex custom agents as standalone TOML role files. Use $subspawn for runtime delegation policy; use this skill to author, validate, install, and test reusable agent roles.
~/.codex/agents.codex/agents under the trusted repotemplates/agents/, a bundled pack, or create anarrow custom role.
scripts/subagent_creator.py for deterministic list, status,plan-sync, install, validate, render, diff, sync, prune, backup, doctor, and smoke workflows.
$subspawn strictrendezvous guidance so spawned results are considered before next work.
Resolve paths relative to this skill directory.
python3 scripts/subagent_creator.py list
python3 scripts/subagent_creator.py list --packs
python3 scripts/subagent_creator.py doctor --project-dir .
python3 scripts/subagent_creator.py status --pack core --project-dir . --include-extra
python3 scripts/subagent_creator.py plan-sync --pack core --target global --include-extra
python3 scripts/subagent_creator.py install reviewer repo_explorer --target global --dry-run
python3 scripts/subagent_creator.py pack install core --target project --project-dir . --dry-run
python3 scripts/subagent_creator.py diff --pack core --target global --include-extra
python3 scripts/subagent_creator.py sync --pack core --target global --dry-run
python3 scripts/subagent_creator.py prune --pack core --target global
python3 scripts/subagent_creator.py backup --target global
python3 scripts/subagent_creator.py validate ~/.codex/agents
python3 scripts/subagent_creator.py smoke --pack docsUse --overwrite only when replacing a role intentionally. Prefer sync when updating already-installed templates because it backs up overwritten files by default. Use --dry-run before writing to global or project config when the target is unclear. Use status --include-extra and plan-sync before broad updates. prune is dry-run only unless --confirm is present.
<name>.toml filenames.gpt-5.5;light read-heavy roles use gpt-5.4-mini.
medium by default, high only for review, security, UI/debug,or complex implementation roles.
default, worker, explorer) unlessthe user explicitly asks for an override. Use names such as repo_explorer.
CI triage roles; workspace-write only for roles that run tests, debug browsers, or implement fixes.
to pin a narrower config.
boundaries, redact secrets, and return evidence in a stable sectioned shape.
core: baseline roles for routine Codex delegation.docs: generic docs, OpenAI docs, Context7, and dependency research.review: PR/code-review helper lanes with false-positive validation.audit: security, runtime, dependency, performance, and docs audit lanes.ops: CI, release, environment, and validation lanes.Use pack list or list --packs to inspect exact membership.
Read references/authoring-guide.md before creating new role families, changing model policy, or adding MCP/server-specific config. Read references/workflow-recipes.md before creating PR review or audit orchestration workflows around the bundled agents.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.