The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell operations including git worktree management (git worktree add), project build and test commands (npm run build, npm test), and local network checks via curl. It also manages a local HTTP server for kanban visualization using python3 -m http.server.
[EXTERNAL_DOWNLOADS]: The skill initiates the download and execution of the @stevestomp/ohno-cli package via npx for task management. This package originates from the skill's author ecosystem.
[PROMPT_INJECTION]: The instructions in references/operating-modes.md and references/bug-fix-pipeline.md direct the agent to bypass standard safety permissions using the --dangerously-skip-permissions flag and bypassPermissions subagent mode. This allows the AI to execute commands without user confirmation in certain operating modes.
[PROMPT_INJECTION]: The skill processes external task data from the ohno MCP which is then used to construct prompts for subagents with shell access, creating a surface for indirect prompt injection.
Ingestion points: Task descriptions and context retrieved via get_task() and get_session_context() in references/subagent-dispatch.md.
Boundary markers: The skill relies on structured tags like [MUST/type] within task descriptions, which provide instructional structure but do not function as security delimiters.
Capability inventory: The skill has capabilities for shell command execution (git, npm, npx), file system modification, and dispatching autonomous subagents.
Sanitization: No explicit process for sanitizing or escaping the content of task descriptions before interpolation is documented.

work-session