dispatching-parallel-agents

Pass

Audited by Gen Agent Trust Hub on Apr 12, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSNO_CODE
Full Analysis
  • [NO_CODE]: The skill consists entirely of Markdown documentation (SKILL.md, ATTRIBUTION.md) and does not include any executable scripts, binaries, or configuration files that run code locally.
  • [EXTERNAL_DOWNLOADS]: The ATTRIBUTION.md file references an external installation source via npx skills add obra/superpowers. This points to a third-party GitHub repository (github.com/obra/superpowers) for acquiring the skill content.
  • [PROMPT_INJECTION]: The skill implements a pattern for 'Indirect Prompt Injection' (Category 8) by instructing agents to ingest untrusted data (test failures and error logs) and interpolate them into prompts for sub-agents.
  • Ingestion points: Test failure messages and test file contents are ingested in SKILL.md (e.g., 'Paste the error messages and test names').
  • Boundary markers: Absent; the instructions suggest direct pasting of external content without delimiters or instructions to ignore embedded commands.
  • Capability inventory: Dispatched agents are given capabilities to read/write production code and execute tests (SKILL.md).
  • Sanitization: Absent; no validation or escaping of the ingested error logs is described.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 12, 2026, 09:36 AM
Security Audit — agent-trust-hub — dispatching-parallel-agents