CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

verify-remote-agent-execution

Check Remote Agent native execution. A stop is not stored as a reply, and a stored reply is not rewritten as a stop. Use when changing initializeRemoteAgentExecutionState, markRemoteAgentExecutionProgress, requestRemoteAgentExecutionCancellation, or completeRemoteAgentExecutionState.

Quality

95%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Verify Remote Agent execution

A stop that wins before a reply is stored never becomes a completed reply, and a reply that is already stored stays a reply. Harm: the Space shows a successful answer after the member stopped the hosted agent, or a stop erases an answer that already landed.

The steps are compare-and-set transitions on one submission, so the spec is TLA+ and the checker is TLC 1.7.4 (jar sha256 936a262061c914694dfd669a543be24573c45d5aa0ff20a8b96b23d01e050e88). The code check calls the real functions with the in-memory storage from services/remote-agent/tests/storage.ts and rereads the terminal outcome using resolveRemoteAgentTerminalExecutionOutcome. Java 11 or newer must be on PATH. The check downloads the jar into ~/.cache/openmeld-tlc/ on first run.

When this applies

Run it when a change touches any of these in services/remote-agent/agent-core/execution-state.ts:

  • initializeRemoteAgentExecutionState
  • markRemoteAgentExecutionProgress
  • requestRemoteAgentExecutionCancellation
  • completeRemoteAgentExecutionState
  • resolveSettledExecutionStatus

Confirmed against commit 965d9eb3219fa2c35789f9ab9a2206ab8699de9b. Move that commit forward only after you have confirmed that RemoteAgentExecution.tla and check.test.ts still describe those functions.

Run

From the repository root, after pnpm install:

node .agents/skills/verify-remote-agent-execution/check.mjs

The stop-then-complete replay must return aborted. The persisted outcome must also be readable as aborted with no completed answer text. A failure during readback catches an inconsistent stored record even if completion returned the expected status.

Read every TLC result with /statechart. The user-facing labels are in view.mjs next to the spec.

What a failure means

  • settles a stop as stopped expected aborted and got completed: resolveSettledExecutionStatus no longer turns cancellation_requested into aborted. That is the bug the rule exists to catch.
  • keeps a stored reply failed: stop after completed changed the row or reported it as stoppable.
  • does not start after a stop that arrived first failed: initialization ignored the admission stop intent.
  • TLC failed while the code check passed: the model allows a step the functions do not, or the opposite. One spec step is one successful compare-and-set. Do not drop AStopIsNotAReplyOrMoreWork to make TLC pass.

Scope

One submission. repairRemoteAgentCompletedExecutionWithoutAnswer is not in the model: a completed row with no stored answer can still become failed. The chart hides the idle end step and a progress tick that stays on Working.

After a change that skipped this skill

Use /maintain-formal-verification-skill to bring this check back in line.

Repository

amplifthq/opentag

Path

.agents/skills/verify-remote-agent-execution/SKILL.md

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.