The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses git clone within scripts/skill-dl to fetch content from GitHub repositories. It also makes network requests to the Serper API (google.serper.dev) for search results and interacts with the npm registry.
[COMMAND_EXECUTION]: The skill's core logic is implemented in bash scripts (scripts/skill-dl, scripts/skill-research.sh), which execute system tools like git, curl, and npx to perform discovery and download tasks.
[DYNAMIC_EXECUTION]: The scripts/skill-dl script utilizes npx -y skills@latest, which dynamically downloads and executes the latest version of the skills utility from the npm registry.
[INDIRECT_PROMPT_INJECTION]: The skill's fundamental purpose is to ingest external, untrusted content. Step 4a in SKILL.md instructs the agent to 'Read the downloaded corpus thoroughly'. This ingestion point represents an attack surface for indirect prompt injection, where malicious instructions embedded in third-party markdown files could attempt to subvert the agent's behavior during the synthesis phase. The skill includes a 'Quality Triage' protocol (Tier 1/2/3) and sanitization logic in scripts/skill-dl (excluding .git and node_modules) to help mitigate risks, but the ingestion surface remains present.

build-skill