The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends the installation of the @membranehq/cli package from the official npm registry. This is a legitimate management tool provided by the vendor to facilitate secure integrations.
[COMMAND_EXECUTION]: The skill utilizes the membrane CLI to perform authentication, search for available actions, and execute Percy workflows. These commands are well-defined and align with the skill's intended purpose of automation.
[CREDENTIALS_UNSAFE]: Authentication is handled through a secure login flow (membrane login) and server-side connections (membrane connect). This approach prevents the exposure of raw secrets or tokens within the agent's context or local file system.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface for indirect injection as it processes data from external Percy actions.
Ingestion points: Outputs from membrane action run and metadata from membrane action list in SKILL.md.
Boundary markers: No specific delimiters are specified for separating untrusted data from the agent's core instructions.
Capability inventory: The skill allows the execution of arbitrary Percy actions via the membrane CLI as described in SKILL.md.
Sanitization: No explicit sanitization or filtering is described; security relies on the agent's built-in guardrails when processing tool outputs.
[PROMPT_INJECTION]: The skill's instructions are focused on functionality and do not contain patterns aimed at overriding safety guidelines or extracting system prompts.

percy