plan-forge

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly requires citing file:line and passing "verbatim code excerpts" into a second-opinion invocation and into emitted implementation prompts, which forces the agent to copy repository content (potentially including API keys/secrets) verbatim into outputs/requests.

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

• Third-party content exposure detected (high risk: 0.85). SKILL.md Step 0 pulls outsider-authored free text from gh issue view ... --json title,body,comments (public/Repo issue content authored by others) and then Step 2/3 uses that context to drive the plan/implementation prompt that is fed into the agent’s LLM context.