The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on shell commands opencode and codex to perform its primary function of delegating work. These commands execute instructions provided as arguments, which are dynamically generated by the agent based on user requests and project context.
[EXTERNAL_DOWNLOADS]: The "Refresh protocol" described in references/models.md instructs the agent to fetch data from https://deepswe.datacurve.ai/ and various provider pricing pages when the internal benchmarking data is considered outdated. This domain is not among the recognized trusted services.
[REMOTE_CODE_EXECUTION]: The core mechanic involves sending prompts to external "executors" (such as Kimi K3 and GPT-5.6) through the opencode and codex interfaces. These executors are instructed to perform tasks like code implementation, refactoring, and test writing, which constitutes remote code execution via the delegation interface.
[DATA_EXFILTRATION]: The skill provides instructions for starting a local server using opencode serve --port <p> and allows other processes to attach to it via http://localhost:<p>. This introduces a potential local network exposure and a vector for unauthorized access to session data.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection via the benchmarking data research protocol.
Ingestion points: Data is ingested from https://deepswe.datacurve.ai/ as specified in references/models.md.
Boundary markers: There are no mentioned boundary markers or "ignore embedded instructions" warnings for this external data.
Capability inventory: The agent utilizes opencode run and codex exec which have the capability to write files, run tests, and execute refactors.
Sanitization: The skill lacks explicit sanitization or validation for the content retrieved from the external benchmarking source before it is used to influence the agent's planning and delegation logic.

use-delegate