The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, untrusted sources including social media (X/Twitter, Reddit) and general web search results.
Ingestion points: External data enters the agent context via browser interactions and search tools (primarily managed through the referenced x9-browser-session skill).
Boundary markers: The skill mandates the use of claim ledgers, explicit labeling of uncertainty, and the preservation of source conflicts. The included validate_html.py script specifically checks for and rejects accidental chat/tool citation markers in the final output.
Capability inventory: The skill possesses file-system write capabilities (creating research dossiers in docs/research/) and local command execution (running the validation script).
Sanitization: The skill employs a deterministic HTML validator (scripts/validate_html.py) that enforces a strict whitelist of allowed HTML tags, blocking dangerous elements like <script>, <iframe>, and <embed>. It also prevents external network dependencies (tracking pixels or external fonts) within inline CSS and ensures all URLs in the report match the verified Markdown source.
[COMMAND_EXECUTION]: The skill instructs the agent to execute a bundled Python script (scripts/validate_html.py) to verify the structural integrity and security of generated HTML reports. This execution is confined to a local script using the system's python3 runtime to perform passive parsing and validation.

x9-research