The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The documentation in SKILL.md suggests that users can provide their API token directly within the chat prompt (e.g., 'token is xxxxx'). This promotes poor security hygiene and risks exposing sensitive credentials in chat logs and conversation histories.- [METADATA_POISONING]: The skill description incorrectly claims to use the 'Python 3.7+ standard library only,' despite all provided code examples in examples.md requiring the third-party requests library. This misleading information can lead to execution errors or unexpected dependency requirements.- [INDIRECT_PROMPT_INJECTION]: The skill fetches data from external, untrusted sources such as issue descriptions, pull request comments, and repository files. This content is interpolated into the agent context without sanitization or boundary markers, which could allow malicious instructions embedded in GitCode content to influence agent behavior.
Ingestion points: Multiple API endpoints defined in reference.md for fetching issues, PRs, commits, and file contents.
Boundary markers: None identified in the instructions for the agent.
Capability inventory: The skill uses the requests library to perform authenticated read and write operations, including creating issues and PRs.
Sanitization: The skill provides no instructions or logic for sanitizing data retrieved from the GitCode API.

gitcode