consensus-summary-html
Pass
Audited by Gen Agent Trust Hub on Sep 8, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script for artifact validation.
- Evidence: The workflow in
SKILL.mdincludes a command to runpython3 "$SKILL_DIR/scripts/validate_consensus_summary.py" <output-file.html>. - [INDIRECT_PROMPT_INJECTION]: The skill ingests external consensus data into an HTML template, which could potentially contain malicious instructions.
- Ingestion points: Reads data from
report_path,state_path, round digests, and seat outputs as specified inSKILL.mdandreferences/input-contract.md. - Boundary markers: Uses
<details>panels and structural placeholders, but lacks explicit sanitization instructions for adversarial input within the consensus data. - Capability inventory: The agent has file system write access and subprocess execution capabilities (running the local validator).
- Sanitization: The
scripts/validate_consensus_summary.pyscript performs structural checks and placeholder detection but does not provide security-focused sanitization of the content.
Audit Metadata