frame-change

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFE
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to execute git rev-parse HEAD to determine the project's current state and whether a feature reconciliation process is required. This is a standard, read-only use of version control tools for environment awareness.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection as it ingests user-provided feature ideas and interacts with repository documentation like CONTEXT.md and docs/specs/INDEX.md. The ingestion points are user prompts and documentation files. The skill's 'Hard Gate' and 6-step checklist act as process boundaries to prevent unintended instruction execution. Capabilities available to the skill include writing to the roadmap and glossary files, and invoking the run-spike sub-skill for code execution.
  • [DYNAMIC_EXECUTION]: The skill involves the use of a run-spike sub-skill to generate and run temporary, throwaway code for fact-checking and prototyping. This is a legitimate functional capability for engineering agents to validate technical assumptions during the design phase.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 03:13 AM
Security Audit — agent-trust-hub — frame-change