The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to execute several shell commands to manage code changes and project state.
Evidence: Use of git pull --rebase, git add ., git commit, bd sync, and git push throughout the protocol steps.
Note: The bd utility is presented as a project-specific tool for backlog synchronization.
[DATA_EXFILTRATION]: The skill performs legitimate network-based data transfers to remote servers.
Evidence: The git push command is explicitly required in the 'Core Principle' and 'Execution Protocol' to share work with the remote repository.
[PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection due to the lack of isolation for external instructions.
Ingestion points: Delegation prompts (Step 1) and repository files (Step 2) serve as primary inputs for agent actions.
Boundary markers: Absent. The skill does not instruct the agent to use delimiters or specific ignore-rules to isolate potentially malicious instructions within the task data.
Capability inventory: The agent is empowered to modify the filesystem, execute git commands, and run the bd synchronization tool.
Sanitization: Absent. The workflow contains no steps for validating or sanitizing the content of external prompts or files before processing them.

subagent