release-notes-narrative

HIGH W007: Insecure credential handling detected in skill instructions.

• Insecure credential handling detected (high risk: 1.00). The prompt explicitly instructs the agent to echo a correlation token embedded in the prompt as the first log line, which forces the LLM to output a verbatim token (a secret-like identifier) and thus requires direct handling/output of sensitive data.

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

• Third-party content exposure detected (high risk: 0.85). The runtime workflow fetches PR body text from GitHub via gh pr view <NNN> --json body --jq .body (Step 6) when the commit body is “thin”; that PR body is outsider-authored free text from non-operating-user contributors and is ingested into the agent/LLM context for narrative synthesis.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill runs gh commands at runtime (e.g., gh pr view and gh release view) against the repository https://github.com/marcusrbrown/systematic to fetch PR/release bodies which are then injected into the model's synthesis process, so external GitHub content directly controls the agent's output.