The Agent Skills Directory

[SAFE]: The skill is primarily instructional, focusing on best practices for documentation architecture and agent behavior engineering. It emphasizes transparency, source-grounding, and rigorous validation.
[PROMPT_INJECTION]: The skill contains explicit safety guidelines to prevent instruction overrides. Specifically, in rules/sourcing.md and rules/forbidden-patterns.md, it instructs the agent to treat retrieved content (web pages, search results) as evidence only and never as instruction authority. This is a strong mitigation against indirect prompt injection.
[EXTERNAL_DOWNLOADS]: The reference files (references/official/anthropic.md and references/official/openai.md) link to official developer documentation from Anthropic and OpenAI. These are well-known technology service providers and are considered safe sources according to standard security practices.
[DATA_EXFILTRATION]: No evidence of credential harvesting or unauthorized data transmission was found. The skill includes instructions in rules/context-engineering.md to define explicit boundaries for data authority and evidence channels, which helps prevent accidental data leakage.
[COMMAND_EXECUTION]: While the skill mentions the use of shell tools for search and file reading/writing, these are standard operations within the intended workflow for document analysis and verification. There are no patterns of arbitrary or dangerous command execution.

docs-maker