The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted external data by extracting text from documents (PDF, Word, spreadsheets). This creates a surface for indirect prompt injection, where malicious instructions hidden in source documents could influence agent behavior during processing or subsequent vault management stages.
Ingestion points: Local and cloud-hosted files such as PDF, DOCX, XLSX, and Markdown as specified in the skill's processing logic.
Boundary markers: The skill does not define specific delimiters to isolate extracted content from the agent's instructions.
Capability inventory: The agent has read/write access to the local file system and can execute scripts for indexing and OCR.
Sanitization: No sanitization or validation of the extracted document text is documented.
[COMMAND_EXECUTION]: The skill utilizes shell commands for its installation and vault health verification processes.
Evidence: Installation instructions involve npx skills add and claude plugin install. The skill also references the execution of vault_index_verify.py for indexing checks.
[EXTERNAL_DOWNLOADS]: The skill downloads necessary plugins and scripts from the developer's official repository.
Evidence: Fetching resources from github.com/soia-team/soia-open-skills and the soia-pkm-vault@soia plugin.

soia-pkm-clip-drive