The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides several bash snippets for the agent to execute using the Bash tool. These include using curl for API requests, jq for JSON processing, and uuidgen for idempotency keys. These commands are necessary for interacting with the document editor's service.- [EXTERNAL_DOWNLOADS]: The skill establishes network connections to www.proofeditor.ai and a local bridge on localhost:9847. These connections are used to fetch document states, read snapshots, and receive updates, which are core features of the skill.- [DATA_EXFILTRATION]: The skill is designed to read the content of local markdown files and transmit them to the external proofeditor.ai service. This is the primary function of the skill (sharing and publishing documents) and is initiated only upon user request or specific workflow triggers. No unauthorized exfiltration of sensitive system credentials or files was observed.- [PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it is designed to ingest and process content from external markdown files.
Ingestion points: The skill reads local markdown files (e.g., plans, drafts) using the Read tool to publish them.
Boundary markers: There are no specific instructions or delimiters provided to the agent to treat the ingested markdown content as untrusted or to ignore any instructions embedded within those files.
Capability inventory: The skill has access to Bash (curl) for network operations, Write for file synchronization, and WebFetch.
Sanitization: The skill utilizes jq --rawfile to safely encode markdown content for JSON transmission, which prevents command injection at the API layer, but it does not sanitize the markdown content itself for potential instruction injection.

ce-proof