The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to process untrusted external data, such as transcripts and webinar archives, to create new agent skills. It contains explicit security instructions in references/content-to-skill-corpus-design.md and references/content-to-skill.md that command the agent to treat such source material strictly as data and to ignore any embedded instructions (e.g., 'ignore previous instructions'). This follows the security best practice of input sanitization and boundary enforcement at the prompt level.
Ingestion points: The skill guides the processing of external transcripts, audio/video content, and newsletter archives.
Boundary markers: The instructions explicitly mandate the use of source identifiers, stable locators, and distinct layers for lessons vs. references.
Capability inventory: The skill does not use dangerous capabilities like shell execution, file writes, or network requests; it is purely instructional and organizational.
Sanitization: The skill provides a protocol for 'Pilot Evaluation' and 'Distribution and Maintenance' which includes manual review of attribution and exactness.
[SAFE]: No other security concerns were identified. There are no hardcoded credentials, remote code execution patterns, persistence mechanisms, or obfuscated content. The static analysis flag for 'prompt injection' was a false positive triggered by these defensive instructions.

lead-magnets