The Agent Skills Directory

[SAFE]: The skill incorporates strong security and privacy practices by explicitly instructing the agent to avoid disclosing sensitive information, such as system prompts, private scratchpads, or credentials, in its output.
[SAFE]: The provided Python script scripts/render_explanation.py serves as a security validator. It implements a function called enforce_content_safety that recursively scans all data fields for forbidden patterns, including private keys, API tokens, and internal model reasoning traces, and rejects the packet if any are found.
[INDIRECT_PROMPT_INJECTION]: The skill processes structured data from PACKET.json which could potentially originate from untrusted external sources.
Ingestion points: Data enters the skill's context through the --input PACKET.json file processed by the render_explanation.py script.
Boundary markers: The skill uses a rigid structure defined in templates/explanation.md with explicit headers to separate different types of evidence and reasoning.
Capability inventory: The skill performs local command execution of its own validation script and chains with documentation and reporting skills.
Sanitization: The skill uses a dedicated validation script that employs regular expressions to detect and block the exfiltration of credentials and private internal deliberation traces.

explain-yourself