The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes the bun runtime to perform a mandatory preflight safety check (preflight.ts) and a release data chunking script. It also utilizes curl for sending batch release data to the configured $RELEASES_API_URL endpoint.
[EXTERNAL_DOWNLOADS]: The agent is instructed to fetch robots.txt and changelog pages from arbitrary user-provided URLs to extract release information. This is part of the skill's primary functionality for changelog onboarding.
[CREDENTIALS_UNSAFE]: The skill utilizes the RELEASES_API_KEY environment variable to authenticate API requests to the backend service. This is documented as standard practice for this local-only developer tool and does not involve hardcoded secrets.
[INDIRECT_PROMPT_INJECTION]: The workflow involves fetching and parsing untrusted data from external websites. While this presents an ingestion surface for potentially malicious instructions hidden in changelog content, the risk is inherent to the skill's primary purpose and is mitigated by the agent's internal safety layers.
Ingestion points: Fetches content from external release pages and sitemaps in Step 3 and Step 4.
Boundary markers: Not explicitly defined in the provided extraction instructions.
Capability inventory: Includes network requests (curl, fetch), local script execution (bun), and sub-agent creation (Agent tool).
Sanitization: Relies on the agent's parsing and structuring of records before transmission.

local-ingest