project-audit

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 0.80). The skill explicitly instructs the agent to "capture key=value outputs" from a discovery script and to save the audit report, and it also reads project/engram artifacts and settings files (which may contain emitted API keys/tokens), so the agent could be required to record secret values verbatim in the generated report/output.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). In PHASE_A_SCRIPT.md, the skill runtime performs an engram lookup via mem_search (e.g., ORPHANED_CHANGES and ENGRAM_HAS_SPECS) that ingests spec/state artifacts stored from prior user/outsider contributions at runtime.