The Agent Skills Directory

[DYNAMIC_CONTEXT_INJECTION]: The skill uses the !command syntax in SKILL.md to execute shell commands at load time. This is used to run a local script (nerf-receipts.js) to generate performance data for the agent's context.
[COMMAND_EXECUTION]: The skill executes node via the shell to run performance tracking scripts. The command paths are relative to the skill directory or plugin root (${CLAUDE_SKILL_DIR} and ${CLAUDE_PLUGIN_ROOT}), which is a standard pattern for integrated tooling.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external script execution and provides instructions on how to interpret that data.
Ingestion points: The output of the nerf-receipts.js script is injected directly into the skill's context.
Boundary markers: None present; the agent is instructed to interpret the injected content directly.
Capability inventory: The skill has shell execution capabilities via the dynamic context injection syntax.
Sanitization: No explicit sanitization or validation of the script output is performed before presentation to the agent.

receipts