The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from multiple external websites and APIs including Celo documentation, MiniPay docs, and DefiLlama. This information is used to update internal markdown reference files and generate pull request descriptions, creating an attack surface where compromised external content could influence the repository.
Ingestion points: The skill fetches content from external sources such as docs.celo.org, api.llama.fi, celopg.eco, and docs.minipay.xyz in SKILL.md (Sources 1-7).
Boundary markers: No explicit delimiters or 'ignore' instructions are used when interpolating external content into repository files or PR descriptions.
Capability inventory: The skill has the ability to write to multiple markdown files (contracts.md, network-info.md, etc.), execute network requests via curl, and perform GitHub CLI operations to create pull requests.
Sanitization: While jq is used for filtering specific API responses, general text-based sanitization for the documentation drift is not explicitly detailed.
[EXTERNAL_DOWNLOADS]: The skill performs numerous network operations to fetch configuration and ecosystem data from external domains.
Evidence: Fetches data from docs.celo.org, api.llama.fi, celopg.eco, docs.minipay.xyz, ramp.ripio.com, and li.quest.
[COMMAND_EXECUTION]: The skill executes shell commands to manage the local environment and interact with the remote repository.
Evidence: Uses git for branch management and gh pr create for opening pull requests against the celo-org/celopedia-skills repository.

docs-watch