The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external sources including social media posts, company news articles, and job descriptions.
Ingestion points: Data is pulled via social_post_list_live, company_enrich, and job_search tools as documented in SKILL.md.
Boundary markers: The instructions do not define strict delimiters or clear "ignore instructions" markers for the external text before it is processed by the model.
Capability inventory: The skill utilizes an execute tool for JavaScript execution and generates CSV files for export.
Sanitization: While the "No-Slop Rule" provides output filtering for AI-typical phrasing, there is no evidence of input sanitization or escaping for the ingested social media or news content.
[DYNAMIC_EXECUTION]: The skill relies on an execute tool to run JavaScript code for orchestrating complex tool calls and data transformations.
Evidence: Multiple JavaScript snippets in SKILL.md demonstrate logic for filtering, mapping, and batching tool responses within the execute environment.
[EXTERNAL_DOWNLOADS]: The skill references and fetches components from external vendor-controlled sources.
Evidence: The README.md instructs users to download a .skill file from the vendor's GitHub releases (github.com/crustdata/skills/releases) and connect to a remote MCP server (install.crustdata.com/mcp). These are official resources belonging to the skill's author.

sales-outreach