The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external project files to populate the dashboard, creating a surface for indirect prompt injection.
Ingestion points: Files located in ~/Projects/ and ~/gsd-workspaces/ under .planning/ directories, specifically HANDOFF.json, STATE.md, ROADMAP.md, and MILESTONES.md (referenced in SKILL.md).
Boundary markers: None are defined; the skill directly interpolates values like next_action text and milestone names into the dashboard output without delimitation.
Capability inventory: The skill has access to Bash for command execution, Read for file access, and WebFetch for network operations.
Sanitization: There is no mention of sanitizing or escaping the text read from project files before it is processed by the agent.
[COMMAND_EXECUTION]: The skill executes a shell script for synchronization tasks.
Evidence: The command bash ${CLAUDE_PLUGIN_ROOT}/scripts/ops-gsd-registry-sync.sh is executed when the user provides the --sync or --refresh arguments. This is a local execution of a script provided by the plugin environment.

ops-projects