The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes a custom command-line interface (wiki) and the bun runtime to perform project memory operations. Specific commands include wiki resume, wiki forge next, wiki search, and synchronization scripts like bun run sync:full.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a retrieval mechanism for an external knowledge vault using wiki ask and wiki search. This creates a standard surface for indirect prompt injection where untrusted data from the vault could potentially influence the agent's behavior.
Ingestion points: Data is ingested via wiki ask and wiki search commands that retrieve content from the directory specified by KNOWLEDGE_VAULT_ROOT.
Boundary markers: The skill mentions a "managed orientation block" in AGENTS.md and CLAUDE.md, which helps differentiate between system-managed instructions and local notes.
Capability inventory: The skill can execute shell commands (wiki, bun) and write to files within the project vault.
Sanitization: The instructions do not explicitly detail sanitization or escaping mechanisms for retrieved content before it is processed by the agent.

wiki