The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple local shell scripts located in ~/.claude/skills/nanostack/bin/ (such as find-artifact.sh, save-solution.sh, and save-artifact.sh). These scripts are used to query previous task data and persist knowledge documents to the local file system.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests and processes data from previous phases which may contain untrusted content.
Ingestion points: Data is read from think, plan, review, qa, security, and ship artifacts using the find-artifact.sh script.
Boundary markers: Absent; the skill does not implement specific delimiters or 'ignore' instructions to isolate artifact data from its operational logic.
Capability inventory: The agent has the ability to execute local shell scripts with arguments and write files to the know-how/solutions/ directory.
Sanitization: Absent; the instructions do not specify any validation, escaping, or filtering for the content retrieved from artifacts before it is used to generate new documents or artifact summaries.

compound