The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes implementation plan data from the local repository (e.g., _index.md, task-NNN-*.md), which is then used to construct prompts for sub-agents spawned via the Agent tool.
Ingestion points: The orchestrator reads task instructions from markdown files and extracts YAML metadata from a plan index file.
Boundary markers: The prompt templates for sub-agents (e.g., in references/batch-execution-playbook.md) do not include explicit instructions or delimiters to isolate untrusted task content from the sub-agent's primary system instructions.
Capability inventory: The skill possesses significant capabilities including the ability to spawn sub-agents, execute shell commands (git, internal scripts), and perform file write/edit operations across the codebase.
Sanitization: There is no evidence of sanitization or filtering of the user-provided plan text before it is interpolated into sub-agent task briefs via lib/task-brief.sh.

executing-plans