append-roadmap-feedback

Pass

Audited by Gen Agent Trust Hub on Jul 18, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests untrusted data from a caller across several fields (Source, Affects, Context, Impact, Recommendation) and writes them to a local file, 'roadmap-feedback.md'. This creates an indirect prompt injection surface.\n
  • Ingestion points: Untrusted data is ingested through the fields specified in the 'Precondition and refusal' section of SKILL.md.\n
  • Boundary markers: The skill uses markdown headers (e.g., ### FBK) to separate entries in SKILL.md, but it lacks specific isolation delimiters for the content within the fields to prevent embedded instructions from being interpreted later.\n
  • Capability inventory: The skill possesses file-write capabilities to the local filesystem as defined in the 'Numbering and append discipline' section of SKILL.md.\n
  • Sanitization: Absent. While there is logic to reject recommendations that assert human decisions, there is no technical sanitization or escaping of the input text itself.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 18, 2026, 01:01 PM
Security Audit — agent-trust-hub — append-roadmap-feedback