The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides installation instructions via npx for the @soia-team/soia-open-skills package. This is a standard deployment mechanism for the skill author's toolset.
[COMMAND_EXECUTION]: The skill instructions involve standard file system operations, such as reading configuration from ~/.config/soia-skills/ and interacting with a local Obsidian vault. These actions are aligned with the skill's stated purpose of personal knowledge management.
[DATA_EXFILTRATION]: The skill accesses personal data within an Obsidian vault, specifically the 个人说明书.md (personal manual) and distilled viewpoints, to inform the writing style and content. This data is used within the agent's context to generate drafts and is not observed being sent to external or unauthorized domains.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect prompt injection as it ingests untrusted data from vault excerpts.
Ingestion points: Reads content from the Obsidian vault (SKILL.md).
Boundary markers: None specified in the instructions to separate user data from agent instructions.
Capability inventory: Writes generated markdown files to the local file system (SKILL.md).
Sanitization: No explicit sanitization or filtering of vault content is described.
While the surface exists, the risk is mitigated by the skill's limited scope (markdown generation) and lack of dangerous tools like arbitrary code execution.

soia-pkm-compose