The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses shell commands like cat, ls, chmod, and find to manage knowledge base files. It also runs local maintenance scripts (scripts/check_env.py and scripts/init_dirs.py) that access local configuration files including openclaw.json in the user's home directory and workspace. It also includes an environment check for the OPENCLAW_WORKSPACE variable to determine platform compatibility.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface as it reads and searches Markdown files from a shared directory without sufficient isolation or instruction filtering.
Ingestion points: Data is ingested from the KNOWLEDGE_BASE_DIR via memory_search and cat kb://... operations as specified in SKILL.md.
Boundary markers: There are no explicit markers or instructions to treat the file content as untrusted data or to ignore embedded instructions during processing.
Capability inventory: The skill possesses the ability to create directories, write files using Python blocks, and modify file permissions via shell commands.
Sanitization: The skill restricts file types to Markdown and limits file sizes to 1 MB, but it does not sanitize or validate the actual text content for malicious directives that might override agent behavior.

claw-knowledge-base