implement-spec

Pass

Audited by Gen Agent Trust Hub on Aug 26, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill explicitly instructs the agent to ignore confirmation prompts and safety-oriented human gates. It uses language such as 'Do not ask for confirmation', 'start executing it immediately', and 'no confirmation before the loop and none between tasks'. This instruction to bypass human oversight is a pattern of agent behavior override.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data in the form of _tasks.md and task_NN.md files which control the execution flow.
  • Ingestion points: Reads graph data and task statuses from Markdown files in docs/specs/ (SKILL.md).
  • Boundary markers: None identified; instructions are interpolated directly from the file content.
  • Capability inventory: The skill has the capability to run other skills (implement-task, qa-gate, archive-spec) which perform file writes, commits, and browser-driven testing (SKILL.md).
  • Sanitization: No evidence of input validation or sanitization for the specification files being parsed.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 26, 2026, 05:20 PM
Security Audit — agent-trust-hub — implement-spec