site-growth-orchestrator
Pass
Audited by Gen Agent Trust Hub on Jul 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill includes instructions to execute a local linter script using
python3 $HOME/SKILL/plans/seo-llm-skill-cluster/scripts/lint_skill_cluster.py $HOME/SKILL. This is a standard validation procedure for the intended orchestration workflow and uses a fixed local path for the script and its arguments. - [PROMPT_INJECTION]: The skill presents an indirect prompt injection surface because it processes external artifacts like YAML and JSON files to coordinate site architecture. However, it implements several mitigating factors:
- Ingestion points: The agent ingests site metadata and planning documents (e.g.,
semantic-core.yaml,url-map.yaml) to track progress and route tasks. - Boundary markers: The skill requires formal handoff packets and evidence-based validation steps (e.g., source-linked notes, audit outputs) before proceeding, which helps maintain context boundaries.
- Capability inventory: The skill can execute local python scripts for linting and perform browser-based verification.
- Sanitization: While explicit sanitization of artifact contents is not detailed, the skill includes strict refusal rules for malicious requests such as hidden bot content or link spam.
Audit Metadata