ワンクリックで
pr-quick-check
Fast security pass over a pull request diff — scanners plus a targeted grep for risky patterns.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
メニュー
Fast security pass over a pull request diff — scanners plus a targeted grep for risky patterns.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
White-box detection of broken authorization (BOLA/IDOR, BFLA, access-control logic flaws) in an API or web codebase. Use when asked to audit authorization, hunt IDOR/BOLA, check object-level or function-level access control, or verify that handlers scope data access to the current principal/tenant. Reconstructs the app's ownership model first, then finds access paths that skip the per-object/per-role check their peers enforce. Static, read-only; delegates injection/secrets/path-traversal to run_semgrep/run_gitleaks.
Find committed secrets and draft a prioritised, step-by-step rotation plan saved to context.
Build a STRIDE threat model of a codebase from its structure and code, using the bundled template.
| name | pr-quick-check |
| description | Fast security pass over a pull request diff — scanners plus a targeted grep for risky patterns. |
| tags | ["pr","review","quick","security"] |
A fast, repeatable security pass over the changes under review. Use it when someone asks for a quick read on a pull request or a set of local changes before a deeper review. It favours signal over completeness: run the scanners, skim for the usual footguns, report what matters.
Frame the change. Use list_files and read_file to understand what
the diff touches — which packages, which entry points, whether secrets,
auth, or input handling are involved. Keep this lightweight; you are
orienting, not auditing every line.
Run the static scanners.
run_semgrep for code-level vulnerability patterns (injection, unsafe
deserialization, weak crypto, path traversal, …).run_gitleaks for committed secrets — API keys, tokens, private keys.Grep for risky patterns the scanners miss. Use grep to look for
anything that warrants a human eye, for example:
TODO|FIXME|HACK|XXX near security-relevant codeeval|exec|os/exec|subprocess (command execution)password|secret|api[_-]?key|token in source (hardcoded credentials)http:// URLs (cleartext transport)InsecureSkipVerify, verify=False)Triage. For each finding, decide: is it introduced by this change, or pre-existing? Is it exploitable, or noise? Prefer reporting a handful of real issues over a long list of low-confidence hits.
Report. Summarise: a one-line verdict (safe to merge / needs changes / blocked), then the findings that drove it, each with file, line, severity, and a concrete fix. Note explicitly if a scanner could not run.