The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses a Python script (scripts/crate_eval.py) to fetch crate metadata and security advisories from external sources including crates.io, docs.rs, GitHub, and RustSec.
Evidence: These are all well-known and trusted services within the software development ecosystem. The script uses standard HTTP libraries and does not download or execute arbitrary code from these locations; it strictly parses metadata (JSON) or scrapes HTML for specific signals (like advisory IDs).
[INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection as it ingests untrusted data from external sources (such as crate descriptions from crates.io) that the AI agent then processes.
Ingestion points: scripts/crate_eval.py fetches text-based metadata from crates.io and GitHub.
Boundary markers: The skill does not explicitly define prompt boundaries for this external data, though it presents it in a structured report format.
Capability inventory: The skill is restricted to read-only network operations and reporting; it does not have capabilities to write files or execute system commands based on the fetched data.
Sanitization: The tool uses standard JSON parsing and regular expressions to extract specific data points, reducing the risk of processing malformed or malicious injection payloads.
[SAFE]: The Python implementation is clean, using only the standard library with no external dependencies. It follows security best practices for registry access, including mandatory User-Agent headers and rate-limit mitigation (request spacing).

rust-crate-discovery