The Agent Skills Directory

[SAFE]: Analysis of the skill instructions and supporting scripts reveals no malicious patterns. The skill is designed for evidence-based research with deterministic validation steps.
[COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/source_evaluator.py, scripts/verify_citations.py, scripts/emit_run_summary.py) for processing research data. These scripts are benign and utilize only standard Python libraries (argparse, json, re, datetime, etc.).
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data fetched from the web.
Ingestion points: External web content retrieved during the investigation phase (P2) described in SKILL.md and references/methodology.md.
Boundary markers: The workflow mandates the creation of structured 'Research Notes' and 'Support Snippets' (defined in references/research-notes-format.md) which requires the agent to parse and rephrase information before it reaches the final synthesis stage.
Capability inventory: The agent can perform web searches, write to a designated workspace/ directory, and execute local validation scripts.
Sanitization: The skill employs scripts/source_evaluator.py to rank sources by domain authority and scripts/verify_citations.py to ensure that citations match the verified source pool, reducing the risk of processing or outputting malicious injected content.

deep-research