The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes multiple bundled bash scripts (preflight.sh, inspect-edge.sh, inspect-network.sh) that invoke system utilities such as docker, git, ip, and ss to verify the deployment environment.
[DYNAMIC_EXECUTION]: The script scripts/inspect-nats.sh generates and executes a Python script via stdin to perform a socket-level NATS handshake. While the script is hardcoded and used for diagnostic purposes, it represents a dynamic execution pattern.
[DATA_EXFILTRATION]: The 'Export logs and learnings' instruction allows the agent to zip site profiles, machine records, and deployment logs to be sent to the user. This legitimate diagnostic feature constitutes a data exposure surface, though the skill explicitly instructs the agent to redact secrets before export.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from external, potentially untrusted command outputs.
Ingestion points: Reads output from docker logs, git status, git ls-remote, and puda machine commands across various reference documents and scripts.
Boundary markers: There are no explicit boundary markers or isolation instructions (e.g., XML tags or delimiters) specified for the interpolation of this external data into the agent's context.
Capability inventory: The skill has access to subprocess execution (bash, python, docker, puda) and file system read/write operations for record-keeping.
Sanitization: There is no evidence of explicit sanitization or filtering of the ingested external command output before it is processed by the agent.

puda-deployment