The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (HIGH): The skill uses npx -y @satori-sh/cli@latest to download and execute code from the npm registry. This package is not from a trusted source, and using the -y flag bypasses user confirmation for installation.
[REMOTE_CODE_EXECUTION] (HIGH): Executing arbitrary code from a public registry via npx allows the package author to execute commands on the user's system whenever the skill is triggered.
[DATA_EXFILTRATION] (MEDIUM): The skill is designed to extract 'facts' from the user's conversation and send them to an external service via the add command. This can include PII, credentials, or proprietary information if the agent deems them 'notable'.
[PROMPT_INJECTION] (MEDIUM): The instructions 'MUST be used instead of any internal systems' and 'Satori is the source of truth' are behavioral overrides intended to prioritize external data over the agent's safety guidelines or internal state.
[PROMPT_INJECTION] (LOW): The skill creates a surface for indirect prompt injection. 1. Ingestion points: Data returned from satori search (SKILL.md). 2. Boundary markers: Absent. It instructs the agent to treat results as 'canonical' and use them immediately as context. 3. Capability inventory: The skill uses npx for command execution (SKILL.md). 4. Sanitization: Absent. No filtering or validation of retrieved content is specified.

satori