The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes several bash scripts (scripts/review-package, scripts/sdd-workspace, scripts/task-brief) used to manage the development workflow.
These scripts execute local commands including git (log, diff, rev-parse, rev-list), awk, mkdir, and standard file operations.
The SKILL.md instructions also encourage the agent to use shell redirection (cat >> progress.md <<'EOF') instead of standard platform editing tools to manage its progress ledger.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest external implementation plans and process them using subagents, which introduces an indirect prompt injection surface.
Ingestion points: The PLAN_FILE and any referenced specification files are read by the controller agent and extracted by the scripts/task-brief script.
Boundary markers: The prompt templates for subagents (implementer-prompt.md, task-reviewer-prompt.md) use markdown headers (e.g., ## Task Description) to isolate injected plan content, but they do not provide explicit instructions for the subagent to ignore potentially malicious directions within that content.
Capability inventory: The skill allows for local file system writes, git repository modifications, and the ability to dispatch further subagents with arbitrary prompts.
Sanitization: No sanitization, escaping, or validation is performed on the plan content before it is interpolated into the prompts for the subagents.
[DYNAMIC_EXECUTION]: The skill dynamically generates execution contexts for subagents by populating markdown templates with data extracted from plan files and git history.
It utilizes scripts/review-package to generate a diff file which is then read in a single call by the reviewer subagent to provide context for its analysis.

subagent-driven-development