The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill fetches metadata and skill source code from a variety of external third-party sources.
Evidence includes fetching marketplace.json and SKILL.md files from domains such as claudemarketplaces.com, skillhub.club, skills.sh, skillsmp.com, glama.ai, pulsemcp.com, and smithery.ai.
It also retrieves raw content from arbitrary GitHub repositories using the raw.githubusercontent.com domain based on search results.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted markdown files and structured metadata from the internet to evaluate them.
Ingestion points: WebFetch is used to retrieve raw SKILL.md content and plugin marketplace JSON files (SKILL.md lines 27-41, reference.md lines 103-108).
Boundary markers: The instructions do not specify any sanitization, escaping, or explicit boundary markers (e.g., XML tags or delimiters) to isolate the untrusted external content during the evaluation process.
Capability inventory: The agent has access to WebSearch, WebFetch, Bash, Read, Glob, and Grep while processing this data.
Sanitization: No sanitization or validation of the fetched content is implemented before it is presented to the model for assessment.

search-skill