The Agent Skills Directory

[COMMAND_EXECUTION]: The skill initialization process uses shell commands such as mkdir -p to create a directory structure for its institutional memory files.
[INDIRECT_PROMPT_INJECTION]: The skill instructions state that constraints recorded in project documentation are binding on all future agent actions, which could allow malicious instructions in those files to influence the agent. 1. Ingestion points: The agent reads instructions from docs/gangsta/constitution.md, docs/gangsta/insights/, and docs/gangsta/fails/ (SKILL.md). 2. Boundary markers: The skill does not define delimiters or provide warnings to the agent to ignore embedded instructions within these data files. 3. Capability inventory: The skill has the ability to create directories, write Markdown files, and read project documentation (SKILL.md). 4. Sanitization: There are no mechanisms described for sanitizing or validating the content of the documentation before it is processed as project commandments or negative constraints.

the-ledger