The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes external, untrusted content which could contain malicious instructions designed to influence the agent's behavior.
Ingestion points: The skill collects 'Source assets' in Phase 1, which includes article text and video transcripts extracted via a local Python script (scripts/video-transcript.py).
Boundary markers: The instructions do not specify the use of delimiters, XML tags, or explicit 'ignore instructions' warnings when handling or interpolating the external content.
Capability inventory: The skill is configured with access to powerful tools, including Bash (used to execute local scripts in Phase 1 and Phase 4), Write (to create idea and draft artifacts), and Read (to load assets and references).
Sanitization: There is no evidence of content sanitization, filtering, or escaping before the external data is processed for idea extraction and drafting.
[COMMAND_EXECUTION]: The skill uses the Bash tool to run local Python scripts for data processing and quality assurance.
Evidence: Phase 1 uses python3 scripts/video-transcript.py for transcript extraction, and Phase 4 executes python3 ~/private-skills/scripts/scan-negative-framing.py to scan drafts for banned phrases.

content-engine