The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It ingests a user-provided domain name and, if enabled, fetches external documentation. This untrusted data is processed by parallel research agents to identify subdomains and propose pipeline steps.
Ingestion points: User prompt (Domain name) and external URLs (via optional Agent 4 Reference Research).
Boundary markers: Absent. No specific delimiters or "ignore embedded instructions" warnings are defined for processing the research content produced by sub-agents.
Capability inventory: File system access (Read/Write), Bash execution, Grep/Glob, and Agent spawning across multiple phases.
Sanitization: Absent. The skill does not describe validation or filtering of the discovered content before it influences the generated Component Manifest and pipeline chains.
[COMMAND_EXECUTION]: The skill uses the Bash tool to perform file system operations.
Evidence: mkdir -p /tmp/pipeline-{run-id}/phase-1-research in Phase 1.
[EXTERNAL_DOWNLOADS]: The 'Reference Research' agent (Agent 4) can be configured to fetch external documentation and specifications from the web to identify domain concepts and validation rules.

domain-research