← All creators

qbs784

User

Stops your coding agent from claiming it verified what it never ran. 17 engineering skills for agent-authored codebases — with the evals, and the retractions, published.

21 indexed · 0 Featured · 0 stars · avg score 73
Prolific

Categories

Indexed Skills (21)

Code & Development Listed

receipts

Shape every answer so a claim about state arrives with the evidence that produced it: name the command, quote its output, say plainly what was not verified. Invoke with /ledger:receipts; stays on until you say "stop receipts".

0 Updated yesterday
qbs784
AI & Automation Listed

using-ledger

The map of this pack — the seven rules, and which discipline to reach for at each stage of a change.

0 Updated yesterday
qbs784
AI & Automation Listed

metabolizing-knowledge

Use when asked whether something delivered earlier still holds, when closing a milestone or running a periodic sweep over what was already marked verified, or when the set of live claims has grown too large to read and dead entries need retiring. DO NOT invoke to add, audit, or retire a decision record — route that to ledger:curating-decision-records. DO NOT invoke to propose removing code — route that to ledger:proving-code-is-dead.

0 Updated yesterday
qbs784
Code & Development Listed

planning-the-work

Use when a plan, unit breakdown, task list, or milestone list is in hand and work is about to start against it — including when asked whether to begin the first unit. Also use when turning an approved approach into units. DO NOT invoke to decide whether the work is worth doing at all — route that to ledger:qualifying-a-request. DO NOT invoke once a unit is underway and the question is whether it is finished — route that to ledger:running-a-bounded-loop.

0 Updated yesterday
qbs784
Code & Development Listed

qualifying-a-request

Use when a request for new work arrives — a feature, a capability, an idea, a leftover being escalated — and before any design, plan, or implementation exists for it. Also use when work already underway turns out to be larger or different than what was agreed. DO NOT invoke to choose which checks to run once the work is underway — route that to ledger:refusing-busywork. DO NOT invoke to work out what a change touches once it is decided — route that to ledger:scoping-a-change.

0 Updated yesterday
qbs784
Code & Development Listed

running-a-bounded-loop

Use when carrying one planned unit of work through to done; when defining what a unit's exit gate has to contain before it can be called finished; or when an unattended run has to decide whether to stop and ask. DO NOT invoke to arrange units, order them, or choose where each runs — route that to ledger:planning-the-work. DO NOT invoke to write the criteria that define done in the first place — route that to ledger:specifying-acceptance.

0 Updated yesterday
qbs784
Code & Development Listed

specifying-acceptance

Use when writing or reviewing acceptance criteria, a definition of done, or the exit conditions for a piece of work; when a requirement reads as an adjective rather than something observable; or when declaring what a component's behavior means. DO NOT invoke to judge whether a guard already written can fail — route that to ledger:proving-the-regression. DO NOT invoke to judge what an already-green signal proves — route that to ledger:what-counts-as-evidence.

0 Updated yesterday
qbs784
Code & Development Listed

validating-real-scenarios

Use when something is being accepted against a real deployment for the first time, when someone reports that delivered behavior is wrong or has drifted, on a periodic health check, or when the question is whether a thing works rather than whether its tests pass. DO NOT invoke to judge what an already-green check proves — route that to ledger:what-counts-as-evidence. DO NOT invoke to carry one unit of work to its exit gate — route that to ledger:running-a-bounded-loop.

0 Updated yesterday
qbs784
Code & Development Listed

adapting-to-a-project

Use when setting up this pack in a repository for the first time, when another skill in this pack reports a missing or unverified adapter value, or after a project's test commands, CI lanes, git hooks, or source layout change. DO NOT invoke to select or run checks for a specific change — route that to ledger:refusing-busywork.

0 Updated yesterday
qbs784
Code & Development Listed

curating-decision-records

Use when adding, auditing, pruning, archiving, restoring, or reviewing durable decision records — architecture decisions, design notes, RFCs, postmortems. Also use when a new record may supersede an existing one, or when the active set has grown hard to read. DO NOT invoke to propose a code removal — route that to ledger:proving-code-is-dead.

0 Updated yesterday
qbs784
Testing & QA Listed

designing-concurrent-tests

Use when writing, changing, or reviewing a test, fixture, or test helper that touches ports, sockets, temporary files, databases, subprocesses, environment variables, the working directory, clocks, timers, global mocks, or asynchronous teardown. DO NOT invoke to investigate an existing intermittent failure — route that to ledger:diagnosing-flakes.

0 Updated yesterday
qbs784
AI & Automation Listed

diagnosing-flakes

Use when a test or CI job fails probabilistically — passes on rerun, fails on one platform only, fails under load, or fails in CI but not locally. Also use when asked to investigate a flaky test, a red run nobody can reproduce, or an intermittent timeout. DO NOT invoke to design a new test for isolation — route that to ledger:designing-concurrent-tests.

0 Updated yesterday
qbs784
Code & Development Listed

fact-checking-by-execution

Use when writing or revising any document, README, comment, or guide that claims a command, a configuration snippet, a default value, an error message, a platform difference, or an install path. Also use when auditing existing documentation for stale claims. DO NOT invoke for prose style or length decisions — route those to ledger:writing-complete-propositions.

0 Updated yesterday
qbs784
AI & Automation Listed

proving-code-is-dead

Use when asked to find things to simplify or delete; when removing a public method, event, config option, package, or test artifact; when replacing hand-rolled code with a dependency; or when auditing defensive copies, validators, freezes, and async lifecycle machinery for speculative complexity. DO NOT invoke to retire a decision record — route that to ledger:curating-decision-records.

0 Updated yesterday
qbs784
Code & Development Listed

proving-the-regression

Use when adding a guard, rule, validator, lint, schema, or check; when fixing a bug and about to assert the fix works; when a test claims to cover a scenario; or before stating that a rule is enforced. DO NOT invoke to choose which checks to run before pushing — route that to ledger:refusing-busywork.

0 Updated yesterday
qbs784
Code & Development Listed

pushing-safely

Use before pushing, force-pushing, or marking a change ready for review; when a push needs history rewritten; when CI reports no checks at all for a pushed commit; or when deciding whether a merge actually landed. DO NOT invoke to select which checks to run first — route that to ledger:refusing-busywork.

0 Updated yesterday
qbs784
Code & Development Listed

recording-ui-evidence

Use when recording a browser or UI interaction demo as a GIF, when a change alters interface behavior a user can see and needs visual evidence, or when attaching a demo GIF to a change request. DO NOT invoke for a single static screenshot — none of this applies to one image.

0 Updated yesterday
qbs784
Code & Development Listed

refusing-busywork

Use when choosing which checks to run before a commit or push; when about to run a full suite or repeat a check that already passed; when trimming prose, deleting code, or archiving records toward a count; or when a refresh, rewrite, or cleanup mechanism exists and you are deciding whether to use it. DO NOT invoke to judge what a passing check proves (ledger:what-counts-as-evidence) or whether an edit is itself right rather than ceremony (ledger:writing-complete-propositions).

0 Updated yesterday
qbs784
Code & Development Listed

scoping-a-change

Use when starting work on a change, before reading or reviewing a diff, or before deciding what a change affects — any time you need to know exactly which files it touches. Also use again after merging or retargeting onto a new base. DO NOT invoke to choose which checks to run over that scope — route that to ledger:refusing-busywork.

0 Updated yesterday
qbs784
Code & Development Listed

what-counts-as-evidence

Use when about to state that something works, passes, is covered, is fixed, is safe, or is done; when reading a green check and deciding what it actually proves; or when tempted to make a check pass by changing the check. DO NOT invoke to choose which checks to run for a change — route that to ledger:refusing-busywork.

0 Updated yesterday
qbs784
Code & Development Listed

writing-complete-propositions

Use when writing, editing, trimming, restoring, or reviewing prose anywhere in a codebase — code comments, docstrings, module headers, READMEs, guides, decision records, prompts, diagnostics, CLI and UI strings — and when deciding whether a passage needs prose at all. DO NOT invoke to hunt reasoning-transcript residue specifically — route that to ledger:trimming-session-vantage.

0 Updated yesterday
qbs784

Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.