spec-engineer

Warn

Audited by Gen Agent Trust Hub on Mar 4, 2026

Risk Level: MEDIUMCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill explicitly generates and facilitates the execution of shell commands. In Phase 3, it provides templates for a 'Self-Fix-Protocol' that includes bash commands for building (npm run build), testing (pytest, playwright), and version control (git add, git commit). Phase 4 further encourages the agent to 'directly get started' with execution upon user approval.
  • [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection (Category 8) because it ingests untrusted data from the local project environment and incorporates it into instructions for downstream agents.
  • Ingestion points: In Phase 1, the skill reads project-level files such as package.json, README.md, CLAUDE.md, and AGENTS.md, as well as the current conversation context.
  • Boundary markers: The skill lacks delimiters or explicit instructions to ignore potentially malicious directions embedded within the ingested project files.
  • Capability inventory: The skill possesses the ability to write files (SPEC.md and prompt files) and provides a structured path for shell command execution via generated bash scripts.
  • Sanitization: No sanitization, escaping, or validation of the content read from the project files is performed before it is interpolated into the generated specifications or agent prompts.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 4, 2026, 04:19 AM