EPLUS Subagent Verification

Plugin eplus-agent-verification from marketplace eplus-verification in Engineering-PLUS/eplus-verification-catalog. Commands read from the repo's manifest — 19 days ago.

Skills (4)

• agent-secret

  Holds the secret word for the eplus-agent-verification preload test. Not for people and not for the main thread; it exists so the preloaded and preloaded-scoped agents can prove whether the skills frontmatter field puts a plugin skill into a subagent's context.

• agent-verification

  Runs the EPLUS subagent behaviour suite inside a live Claude desktop Chat session and writes a verdict report. Use when the user runs /eplus-agent-verification:run, asks to "test the subagents", "check what background agents do here", or wants proof of how plugin agents behave on this seat before a

• fork-bg

  Test skill for eplus-agent-verification that runs its body in the echo agent through context fork with the default background setting, so the result is supposed to arrive later. Invoke it only during an agent-verification run, with the run token as the argument.

• fork-wait

  Test skill for eplus-agent-verification that runs its body in the echo agent through context fork with background false, so the turn is supposed to wait for the result. Invoke it only during an agent-verification run, with the run token as the argument.

Slash commands (2)

• /eplus-agent-verification:probe

  Zero-cost inventory for eplus-agent-verification. Reports the session environment, whether the Agent tool exists on this seat, which eplus-agent-verification agents are listed, which messaging tools exist, and the Agent tool's parameter names. No tools, no spawns.

• /eplus-agent-verification:run

  Run the EPLUS subagent behaviour suite in this Chat session (plain, background and skill-forked spawns of the plugin's own Haiku probes, preload, model, maxTurns, nested spawn, memory, hand-back channel) and write the verdict report. Arguments - all | --quick | <card number> [--token <word>]

Subagents (9)

• capped

  Test worker for eplus-agent-verification with maxTurns 2, to prove the turn cap and the partial-result marker, and to give the main thread something to resume with SendMessage. Spawn it only when the agent-verification skill says so. Runs six one-per-turn shell echoes and returns one CAPPED line.

• echo-bg

  Test worker for eplus-agent-verification, identical to echo but declared background true in its frontmatter. Spawn it only when the agent-verification skill says so. Returns one ECHO-BG line with the token, the model name, whether SubagentHandback is among its tools, and its tool names.

• echo

  Test worker for eplus-agent-verification. Spawn it only when the agent-verification skill or one of its commands says so; it does no real work. Returns one ECHO line carrying the token it was given, the model name its instructions show, whether SubagentHandback is among its tools, and its tool names

• nameless

  Test agent for eplus-agent-verification with no name field, to see how the loader registers it (the plugins reference says a plugin agent without a name is named after its file, so this should appear as eplus-agent-verification:nameless). Presence check only; never spawn it.

• preloaded-scoped

  Test worker for eplus-agent-verification whose frontmatter preloads the agent-secret skill by its plugin-scoped name. Spawn it only when the agent-verification skill says so. Returns one PRELOAD line saying whether the secret word reached its context.

• preloaded

  Test worker for eplus-agent-verification whose frontmatter preloads the agent-secret skill by its bare name. Spawn it only when the agent-verification skill says so. Returns one PRELOAD line saying whether the secret word reached its context.

• relay

  Test worker for eplus-agent-verification that spawns the echo worker itself, to prove nested subagent spawning. Spawn it only when the agent-verification skill says so. Returns one RELAY line carrying the child's reply or the exact error.

• rememberer

  Test worker for eplus-agent-verification with memory scope project, to prove persistent subagent memory across spawns. Spawn it only when the agent-verification skill says so, always twice in a row with the same token. Returns one MEMORY line naming the previous run token it found, then records this

• sonnet-probe

  Test worker for eplus-agent-verification that runs on Sonnet at low effort, to prove the model and effort frontmatter fields are honoured. Spawn it only when the agent-verification skill says so. Returns one PROBE line with the token, the model name, whether SubagentHandback is among its tools, and

Hooks (4)

• PreToolUse

  & "${CLAUDE_PLUGIN_ROOT}\scripts\agent-probe.ps1"

• PostToolUse

  & "${CLAUDE_PLUGIN_ROOT}\scripts\agent-probe.ps1"

• SubagentStart

  & "${CLAUDE_PLUGIN_ROOT}\scripts\agent-probe.ps1"

• SubagentStop

  & "${CLAUDE_PLUGIN_ROOT}\scripts\agent-probe.ps1"

Similar plugins