sg-mission-lock

Pass

Audited by Gen Agent Trust Hub on Sep 4, 2026

Risk Level: SAFE
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill implements a robust defense against indirect prompt injection by explicitly instructing the agent to treat all data retrieved during a mission (e.g., DOM snapshots, logs, files) as evidence only, never as instructions. It specifically identifies that embedded directives in read content should be reported as findings rather than followed.
  • [PROMPT_INJECTION]: The file references/scenarios.md contains examples of prompt injection strings such as "ignore previous instructions." These are clearly documented as regression test cases to verify the agent's ability to resist such attacks, rather than being an attempt by the skill to bypass security guidelines.
  • [COMMAND_EXECUTION]: The skill contains local Node.js scripts (scripts/inject-mission-lock.mjs and scripts/mission-lock-smoke-test.mjs) used for platform hook integration and automated testing. These scripts use standard Node.js built-ins to process context and do not perform unauthorized network operations or file system mutations.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 4, 2026, 08:34 AM
Security Audit — agent-trust-hub — sg-mission-lock