The Agent Skills Directory

[SAFE]: The skill provides legitimate procedural guidelines for managing a knowledge base and integrating it into an agent's development loop.\n- [EXTERNAL_DOWNLOADS]: The skill references installing the skillwiki CLI via npm and syncing with a Git remote. These are standard operations for the described tool and align with the author's namespace (karlorz/wiki).\n- [DATA_EXFILTRATION]: The skill contains a strict Sensitive Content Policy that mandates redacting credentials before they are stored in the vault, effectively reducing the risk of accidental exposure of secrets.\n- [PROMPT_INJECTION]: The skill manages a workflow for ingesting untrusted external data (via wiki-ingest and wiki-add-task) into a knowledge base that is later queried by the agent. This represents an indirect prompt injection surface.\n
Ingestion points: wiki-ingest (URL/text conversion) and wiki-add-task (ad-hoc captures).\n
Boundary markers: The skill uses architectural layers (raw/, projects/) and publication contracts but lacks specific mention of instruction-isolation delimiters for ingested text.\n
Capability inventory: skillwiki CLI, git, and gh for file modification and synchronization.\n
Sanitization: A strict credential redaction policy is enforced for all ingested sources.

using-skillwiki