The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-supplied data, including post briefs in JSON format and draft text, to generate recommendations and audit content quality. This ingestion is a standard attack surface for indirect prompt injection.
Ingestion points: User-provided inputs for the angle_picker.py and post_gate.py scripts, as well as the initial post brief.
Boundary markers: The instructions include a manual verification checklist in references/pre-publish-gate.md and a mechanical gate script to identify placeholders and stock phrases.
Capability inventory: The skill is restricted to text analysis and console output. It does not perform network operations, file system writes (outside of standard output), or high-privilege system commands.
Sanitization: No specific sanitization or escaping of user-provided material is implemented in the scripts before processing.
[COMMAND_EXECUTION]: The skill utilizes local Python scripts (scripts/angle_picker.py and scripts/post_gate.py) to perform its core functions.
The scripts use standard Python libraries (argparse, json, re, pathlib) for text processing and linting.
Analysis confirms that the scripts do not spawn shell processes, execute arbitrary code, or access sensitive environment variables/files.

linkedin-post-writer