chore(security): pin claude-code-action to known-good v1.0.174

[abir-vim]

Automated fleet sync of the org security scanner. Pins claude-code-action off the floating @v1 tag (which regressed at v1.0.175+ and silently no-ops on Bedrock, failing the scan gate closed) to the known-good SHA. See scanner-org for the source template.

[github-actions]

Security Scanner Results

Critical	High	Medium	Low	Info	Dep BLOCK	Dep WARN
0	0	1	1	0	0	0

Findings

• 🟡 allowed_bots: "*" disables the bot-author guard for every bot on all scan attempts — .github/workflows/security.yml:170
  • Remediation: Replace the wildcard with an explicit list of the trusted automation accounts actually needed, e.g. allowed_bots: "renovate[bot],dependabot[bot],gitops[bot]" (match the exact bot login your automation uses). This keeps the intended bot PRs covered while denying the guard to any other bot identity.
• 🔵 Enabling bot triggers extends the prompt-injection / gate-evasion surface of the AI scanner — .github/workflows/security.yml:167
  • Remediation: Scope allowed_bots to specific trusted bots (see the wildcard-bot-trigger-allowlist finding). Additionally, treat the AI findings.json as advisory-only for gating and keep the deterministic OSV/dependency gate as the hard block, or add an out-of-band integrity check (e.g. require the phases-completed check plus a minimum-analysis sanity signal) so an injected empty result cannot silently pass the gate.

Gate blocks at severity critical or a dependency marked BLOCK. Full history in scanner-stats.