Skip to content

xiduzo/wtf / skills/wtf.write-trace

wtf.write-trace

Verified

This skill should be used when a user wants to create a single Trace issue under a Feature — for example "create a trace", "write a trace for this feature", "add a trace to feature

evals/evals.json

View on GitHub
{
  "skill_name": "wtf.write-trace",
  "evals": [
    {
      "id": 1,
      "prompt": "Run wtf.write-trace in test mode. Skip steps 0–1 (gh setup confirmed; instead of gh commands, read `evals/fixtures/claim-selection.json` and use its `feature` and `traces` objects as the live data). The confirmed claim is Trace Plan entry 2 of Feature #55. Produce the Story and Scenario Claim sections of the Trace body as step 7 would draft them. Do not produce other sections. Do not create an issue.",
      "expected_output": "The Story section copies the Merchant PDF-export story verbatim from the Feature. The Scenario Claim lists exactly the three scenario names from plan entry 2, plus a collapsed <details> block whose summary reads 'synced from Feature #55 — edit there, not here' and whose Gherkin is copied verbatim from the Feature body. The Skeleton's claimed scenario ('Merchant downloads a settlement PDF') is not claimed again, and no new scenarios are invented.",
      "files": ["evals/fixtures/claim-selection.json"],
      "expectations": [
        "The Story section reproduces the Feature's story text verbatim (Merchant, export settlement history as a PDF, accounting and audit needs) without rewording",
        "The Scenario Claim names exactly the three scenarios from plan entry 2: 'PDF header shows Merchant name and date range', 'Empty settlement range produces a placeholder PDF', and 'Merchant without export permission sees no export action'",
        "The claim does not include 'Merchant downloads a settlement PDF' (already claimed by existing Trace #56)",
        "A <details> block is present whose summary contains 'synced from Feature #55' and 'edit there, not here'",
        "The Gherkin inside the <details> block matches the Feature body's scenarios verbatim — no re-derived or newly invented steps or scenarios"
      ]
    },
    {
      "id": 2,
      "prompt": "Run wtf.write-trace in test mode. Skip steps 0–7 (all steps before the scope gate). The Trace draft is provided in `evals/fixtures/depth-split-draft.json`, field `trace.draft_body`, with the parent Feature in field `feature`. Run step 8 only: the scope gate. Output whether a split is needed, which signals fire, and what concrete split is proposed.",
      "expected_output": "Step 8 fires the claim-too-large signal: six scenarios spanning distinct failure domains is too much for one agent pass, and a Skeleton must claim the happy path minimally. It proposes a depth split — a re-partitioned Scenario Claim, e.g. Skeleton claims 'Merchant downloads a settlement PDF' and Deepening Traces claim the header, empty-range, permission, and slow-export scenarios. It never proposes a layer split, and it presents the keep/split/stop choice.",
      "files": ["evals/fixtures/depth-split-draft.json"],
      "expectations": [
        "At least one split signal is identified (the Scenario Claim is too large for one agent pass, or the Skeleton claims more than the happy path)",
        "The proposed split re-partitions the Scenario Claim by depth: a smaller first claim (the happy-path scenario) plus one or more Deepening Traces claiming the remaining scenarios",
        "No layer split is proposed — the output does not split the work into model/schema, API/backend, and UI/frontend units",
        "The keep/split/stop choice is presented to the user",
        "The draft is not silently approved without raising the split question"
      ]
    },
    {
      "id": 3,
      "prompt": "Run wtf.write-trace in test mode. Skip steps 0–1 (gh setup confirmed; instead of gh commands, read `evals/fixtures/ambiguous-claim.json` and use its `feature` and `traces` objects as the live data). The user request is in field `trace_request`: 'Add a trace for settlement error handling'. Run step 2 only: select the story and the Scenario Claim. Show what you would present and what you would ask. Do not answer for the user and do not proceed to step 3.",
      "expected_output": "Step 2 shows the claim state — 'Failed settlement is retried' is claimed by Trace #61, four scenarios remain unclaimed, and the Trace Plan has no remaining entries — then asks the user which of the unclaimed scenarios this Trace claims. It does not invent a claim, does not re-claim the Skeleton's scenario, and does not draft a body.",
      "files": ["evals/fixtures/ambiguous-claim.json"],
      "expectations": [
        "The user is asked which scenarios to claim rather than a claim being assumed from the vague request",
        "The four unclaimed scenarios (permanent failure, Merchant notification, operator retry, idempotent duplicate retry) are surfaced as the selectable claim space",
        "The already-claimed scenario 'Failed settlement is retried' is shown as taken by Trace #61 and is not offered for claiming",
        "The output does not proceed to later steps or produce a Trace body draft"
      ]
    }
  ]
}
reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis.Stars: 471kUpdated Claude CodeCodexVerifiedInstall
diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness.Stars: 322kUpdated Claude CodeCodexVerifiedInstall
dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependenciesStars: 322kUpdated Claude CodeCodexVerifiedInstall
executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is availableStars: 322kUpdated Claude CodeCodexVerifiedInstall