The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the execution of a local Python script (scripts/plan_graph.py) to perform file system operations. This script includes capabilities to create, modify, and delete files (specifically .plan.md files and JSON state snapshots) on the host system.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes untrusted data from .plan.md files.
Ingestion points: The script scripts/plan_graph.py reads local files with the .plan.md extension, parsing their YAML frontmatter and presenting the contents to the agent context.
Boundary markers: The skill lacks explicit delimiters or instructions (e.g., 'ignore embedded instructions') to prevent the agent from accidentally executing instructions found within the data fields of the plan files.
Capability inventory: The agent has access to file writing and deletion tools via the generate and delete-plan commands in the provided Python script.
Sanitization: The custom frontmatter parser in scripts/plan_graph.py does not filter or sanitize the values extracted from the todos or overview fields for potential malicious instructions.

plan-graph