The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and parses YAML and JSON 'Seed' files from the local filesystem (~/.ouroboros/seeds/). The data from these files is interpolated directly into the titles and bodies of GitHub issues.
Ingestion points: Local specification files identified in Step 2 and parsed in Step 3.
Boundary markers: The skill uses quoted heredocs (--body "$(cat <<'BODY' ... BODY)") in Step 7, which effectively prevents shell command injection from the file content into the execution environment.
Capability inventory: The skill uses run_shell to execute gh CLI commands and perform file system operations.
Sanitization: While shell-level injection is prevented, the skill does not explicitly sanitize the natural language content within the seed files before it is processed by the agent to create issue descriptions.
[COMMAND_EXECUTION]: The skill utilizes the gh (GitHub) CLI to perform repository operations, including creating labels, creating issues, and adding comments.
Target Validation: The skill explicitly asks the user for the target repository (TARGET_REPO) before performing write operations.
Tool Usage: Commands such as gh label create, gh issue create, and gh issue comment are used according to standard development workflows.

publish