The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and processes content from the external Pidj platform.
Ingestion points: Untrusted data from Pidj is ingested into the agent context via the output of commands like membrane action list and membrane action run (identified in SKILL.md).
Boundary markers: No explicit delimiters or instructions are provided to the agent to treat external content as potentially untrusted data.
Capability inventory: The agent can execute platform actions and shell commands via the membrane CLI (identified in SKILL.md).
Sanitization: The instructions do not include mechanisms for sanitizing or validating retrieved data before processing.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download and installation of a command-line utility from the NPM registry.
Evidence: The command npm install -g @membranehq/cli@latest is used to install the required vendor-provided CLI tool (found in SKILL.md).
[COMMAND_EXECUTION]: The skill's operational logic depends on the execution of shell commands through a CLI interface.
Evidence: The documentation provides numerous shell command examples including membrane login, membrane action list, and membrane action run (found in SKILL.md).

pidj