The Agent Skills Directory

[PROMPT_INJECTION]: The skill instructs the agent to enter an "autonomous mode" once a work plan is provided, stating that the initial instruction serves as "batch approval" for all subsequent subagent calls and tasks. This directive aims to bypass per-step user confirmation.
[INDIRECT_PROMPT_INJECTION]: The skill's workflow is driven by the content of markdown files located in docs/plans/ and docs/plans/tasks/. These files are ingested into the agent's context and used to determine which subagents to call and what actions to take. * Ingestion points: docs/plans/*.md, docs/plans/tasks/*.md, and user-provided $ARGUMENTS. * Boundary markers: No explicit delimiters or instructions to ignore embedded commands within the ingested files are defined. * Capability inventory: The orchestrator can invoke various specialized subagents (task-executor, quality-fixer, code-reviewer, security-reviewer), execute git commit commands, and delete files from the local filesystem. * Sanitization: There is no evidence of sanitization or validation of the content within the ingested task files or work plans.
[COMMAND_EXECUTION]: The skill performs automated filesystem cleanup by deleting task files and executes git commit after tasks are completed, following a structured 4-step execution cycle.

recipe-fullstack-build