mission-metrics

Pass

Audited by Gen Agent Trust Hub on Sep 20, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites via research tasks and from session logs through session mining.
  • Ingestion points: Web research results (WebFetch) and session history mining are used to shape the interview questions.
  • Boundary markers: Absent. The instructions do not direct the agent to delimit or sanitize external content, nor to ignore potential instructions embedded within that data.
  • Capability inventory: The agent can read and write files in the repository and the user's home directory, and perform network requests via WebFetch.
  • Sanitization: Absent.
  • [DATA_EXPOSURE]: The skill interacts with files outside the repository's immediate scope, specifically targeting the user's personal vault at ~/vault/missions.md and related specs. While central to the skill's purpose of long-term mission tracking, this grants the agent access to potentially sensitive personal data in the user's home directory.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 20, 2026, 08:52 PM
Security Audit — agent-trust-hub — mission-metrics