The Agent Skills Directory

[PROMPT_INJECTION]: The skill documentation facilitates the ingestion of untrusted data from Gitea instances, which presents a surface for potential indirect prompt injection attacks.
Ingestion points: The agent can retrieve external content such as issue descriptions, pull request comments, and workflow logs through commands like tea issues list, tea pulls list, and tea actions runs logs (documented in files like references/commands/issues.md and references/commands/pulls.md).
Boundary markers: No explicit instructions or delimiters are provided to help the agent distinguish between its system instructions and the content retrieved from external sources.
Capability inventory: The skill documents capabilities for significant modifications to Gitea instances, including merging pull requests, managing repository secrets (tea actions secrets create), and performing administrative tasks such as creating users with administrator privileges (tea admin users create --admin).
Sanitization: There is no mention of sanitization or validation mechanisms for content fetched from external Gitea instances.
[COMMAND_EXECUTION]: The reference files detail various CLI commands for managing Gitea entities, including administrative operations and credential management. It covers creating users with administrative rights and handling access tokens for authentication, which are standard functionalities of the administrative toolset.
[REMOTE_CODE_EXECUTION]: The documentation includes standard procedures for tool installation using official package managers and Go's module system (go install gitea.dev/tea). It also outlines the standard method for enabling shell completions by sourcing scripts generated at runtime (source <(tea completion ...)).

tea-cli