The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes untrusted data from LinkedIn profiles, including post text, comments, and headlines.
Ingestion points: Data is retrieved from external LinkedIn profiles via the person_enrich and social_post_list_live tools.
Boundary markers: The skill contains multiple explicit instructions to the agent to treat external content solely as data and never as instructions, found in SKILL.md, references/angles.md, and references/output-contract.md.
Capability inventory: The skill's capabilities are limited to writing local data files (.json, .md, .xlsx) and calling defined API tools; it does not perform dynamic code evaluation on ingested data.
Sanitization: The merge.py and build_sheet.py scripts implement sanitization logic to remove surrogate characters and control characters from data before it is written to the final reports or Excel workbooks.
[COMMAND_EXECUTION]: The test suite evals/test_merge.py uses subprocess.run to execute the skill's own internal processing scripts (merge.py, shard_resolve.py, and build_sheet.py) for validation purposes using the current Python interpreter.
[EXTERNAL_DOWNLOADS]: The documentation and scripts reference the installation of the openpyxl Python library from standard package registries to facilitate Excel file generation.
[DATA_EXPOSURE]: The skill processes professional network data utilizing the vendor's API infrastructure (crustdata.com) and includes logic to respect user privacy settings by handling redacted profile statuses.

social-connections