The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses local shell commands to manage its workflow, specifically git status --porcelain to ensure the workspace is clean before applying changes and uvx skilllint@latest to perform automated checks on the modified skill files.
[EXTERNAL_DOWNLOADS]: To verify if a skill is out of date, this skill fetches content from external documentation URLs. According to references/url-fetch-spec.md, it utilizes a fallback chain that includes using web search tools and cloning GitHub repositories (git clone) to locate and mine documentation files if direct access is blocked.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest data from external, untrusted sources (web docs and repositories) and use that data to automatically generate a change plan and update skill instructions.
Ingestion points: The skill extracts documentation URLs from SKILL.md and discovers new ones via search engines during the drift detection stage.
Boundary markers: The provided files do not specify the use of clear delimiters or instructions that tell the sub-agents to ignore potentially malicious directions embedded within the fetched documentation content.
Capability inventory: The skill possesses significant capabilities, including network access, repository cloning, execution of tools via uvx, and the ability to write updates directly to the local file system.
Sanitization: There is no evidence of content sanitization or validation routines to filter out instruction-like patterns from the ingested documentation before they reach the synthesis and write stages.

skill-sync