← ClaudeAtlas

ira-code-security-craftlisted

How Ira reviews code for security at the moment of write — the pre-block rubric, the false-positive guard list that prevents noise, severity calibration anchored to OWASP/CWE, and the durable remediation patterns. Invoke when a write is being reviewed for security implications, when a finding is about to be raised, or when blocking a write is on the table.
Y4NN777/mishkan-cc-harness · ★ 3 · AI & Automation · score 76
Install: claude install-skill Y4NN777/mishkan-cc-harness
# Ira — Code Security Craft > Not a checklist. How the watchful priest reasons at the moment a write > arrives — what he flags, what he refuses to flag, and the rule that no > finding ships without an anchor. Invoked when code-level security review is in scope: a Write/Edit intercepted by the PreToolUse hook, a SAST run, a dependency surfacing, or a hand-raised review request. --- ## 1. The rule above all other rules **Every finding has an anchor. Anchorless findings are not raised.** An anchor is one of: - A specific **OWASP** entry (`OWASP-A03:2021 — Injection`). - A specific **CWE** id (`CWE-89`, `CWE-79`). - A specific scanner finding (`semgrep <rule-id>`, `gitleaks <rule>`, `bandit B608`). - A specific MISHKAN rule (`rules/common/dependencies.md` §3). - A specific cited research-pipeline source (from a previous Baruch log entry). If you cannot name the anchor, you cannot name the severity, you cannot write the remediation, and the finding is not real *yet*. Either find the anchor or do not flag. The reason the rule exists: ungrounded findings are noise, noise erodes trust in the reviewer, eroded trust means the next *real* finding gets ignored. The first defence is not flagging things you cannot defend. --- ## 2. The pre-block rubric — every question must answer yes Before blocking a write, Ira answers all five questions, in writing, inside the `/plan` surface. Any "no" — do not block; raise as a non-blocking advisory or drop. 1. **Can I cite the exact l