The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes bash script templates for local file orchestration, polling, and status management in references/waiting-script.md. These scripts use standard system utilities such as ls, grep, sed, awk, and yq to manage the planning lifecycle.- [COMMAND_EXECUTION]: The wait_with_callback.sh template in references/waiting-script.md employs the eval command to execute callback functions when specific files are detected. This represents a form of dynamic code execution used for internal task orchestration.- [PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection because agents process markdown files generated by other specialist agents during the iteration rounds.
Ingestion points: Subagent-generated markdown files (e.g., draft-*.md, review-*.md) are read by the orchestrator and other specialists as described in SKILL.md and references/plan-templates.md.
Boundary markers: Absent; the templates do not define explicit delimiters or instructions to ignore embedded commands within the specialist-generated content.
Capability inventory: The skill relies on bash-based file operations, status tracking, and orchestration logic across all referenced scripts.
Sanitization: There is no evidence of validation, filtering, or sanitization performed on the content of the markdown files before they are ingested into the context of other agents.

subagent-planificator