The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (scan.sh, quick-diff.sh, save-results.sh) to perform file system inventory, compare file modification times, and manage evaluation results. These scripts use system utilities like find, awk, and jq to process skill metadata and usage logs.
[INDIRECT_PROMPT_INJECTION]: The skill ingests and processes content from other skills which may contain untrusted data, presenting a surface for indirect prompt injection attacks where malicious instructions in an audited skill could influence the subagent's behavior.
Ingestion points: The skill reads all SKILL.md files located in the ~/.claude/skills/ directory and any project-specific skill directories.
Boundary markers: The subagent prompt uses [INVENTORY] and [CHECKLIST] headers to delimit data, but lacks explicit instructions or guardrails to ignore potentially malicious commands embedded within the skill files being analyzed.
Capability inventory: The skill uses a general-purpose subagent to evaluate quality and recommends actions like 'Retire', 'Merge', or 'Improve', which are then presented to the user for confirmation and execution.
Sanitization: There is no evidence of sanitization, filtering, or escaping of the content extracted from the audited skills before it is interpolated into the subagent's prompt.

skill-stocktake