verification plugins

by Thomas Loosch

The portable half of a solo agentic engineering harness: the skills pipeline, the verification checklist, the AI-failure-mode code review checklist, and the catch-log schema. Mechanical wiring (CI, hooks, lint) is installed per-project by bootstrap, not by this plugin.

by Luke Awino

Agentic workflow orchestration with explicit plans, evidence, and completion gates.

by Khalil Maaouni

A senior backend and data engineering colleague. Designs systems in the order the work runs (purpose, process, architecture, data, expression, verification), produces a design dossier sized by a scored intake, and holds the result to mechanical gates where absent evidence is NO-DATA and never a pass.

by Bymax One

Real-browser verification and assisted UI testing for web apps. /bymax-web-verify:test brings the full stack up — reuses a running backend instance or starts one, opens the frontend in the Claude Desktop Browser pane (preview) — and walks a UI flow step by step while the user watches, verifying via page state, console, network (2xx gate), and server logs; falls back to the agent-browser CLI in a terminal. /bymax-web-verify:verify drives a real browser to confirm a single change works. /bymax-web-verify:setup installs the agent-browser CLI + Chrome for Testing in one shot. A SessionStart hook detects a missing CLI and offers to install it. Depends on `agent-browser` but never bundles it. /bymax-web-verify:record captures a UI flow as reviewable video evidence: the project's own Playwright with video forced on for one isolated run, human pacing (slowMo with a resized timeout), ffmpeg lead-in trim with a verified first frame, an ffprobe pacing gate, optional --mp4, and a plain-text walkthrough derived from the spec that ran.

by SriBYJU

Adaptive model, reasoning, context, verification, and parallel-work orchestration for Codex.

by HernandoR

Git workflow, versioning, worktrees, Justfile recipes, TDD, test shrinking, and best practices

by Thomas Fournier

Independent causal verification for code produced by Codex and other coding agents.

by yukineko

v0.1.30 integrates this line with the remote's: the session-scoped skip now carries a third state. Session-scoping fixed WHOSE token is consumed; it did not fix WHEN. A Stop is adjudicated by four independent gate processes, so a token honoured by one that allows was burned even when another blocked the same stop — the stop never happened and the one-shot escape had been spent on nothing, which is standing pressure toward a permanent bypass (CLAUDE.md 5). `consume_session_skip` now takes `stop_hook_active` and parks an honoured token as `<id>.skip.honoured`: it is re-honoured while a re-entry after a block is in flight, and deleted the first time a chain actually ends. A rename failure that can neither bound nor clear the token is resolved to NOT honouring it. v0.1.28: the shared project-root `.donegate-skip` marker is gone. The operator escape is now `donegate skip --reason "..."`, which is scoped to the issuing session (CLAUDE.md 5 forbids a shared one-shot file under parallel sessions, since whichever session stops next consumes it), requires a non-empty reason, and appends both the issue and the consumption to the gate log so a bypass cannot happen unrecorded. v0.1.21: re-blesses the verdict compile-fail trybuild golden stderr again — CI's `dtolnay/rust-toolchain@stable` installs rustc without the `rust-src` component, which omits the multi-line snippet in `note: the method ... exists on the type ...` diagnostics; a local toolchain with `rust-src` installed renders the longer form and silently re-diverges the golden. Reproduced by installing a `--profile minimal` (no rust-src) toolchain matching CI exactly. No contract change. v0.1.20: re-blesses the verdict compile-fail trybuild golden stderr (rustc note-formatting drift only, no contract change). Completion-verification gate for Claude Code: on Stop, run your acceptance commands as subprocesses and block the stop until every required check is green, feeding failures back for an auto-fix loop. Subscription-native (one hook + bundled Rust binary, no API key).

by DSH AI Software Workflow

A ten-skill development workflow for AI software projects, distributed as a Codex plugin for product confirmation, planning, implementation, verification, repair, review, delegation, research, and release.

by claude-skills

Routes work from an expensive lead model to tool-scoped worker agents -- explorer and researcher on Haiku, editor and test-runner on Sonnet, refuter and investigator on Opus -- and makes the lead re-verify what they report instead of trusting it. Adds fabflows:build, a deterministic loop in which an Opus builder implements a spec and a fresh reviewer returns ACCEPT or REWORK. Ships an ACTIVE guard hook that blocks package installs, commits and pushes to a default branch, destructive shell commands, and reads or writes of credential files; the guard fails open, so a bug degrades to no guard rather than a blocked session.

by Jev-decision

Build AI-powered software with TypeSafe System One and Jev using typed judgments, calibrated probabilities, and composable decision primitives.

by Jev-decision

Build AI-powered software with TypeSafe System One and Jev using typed judgments, calibrated probabilities, and composable decision primitives.

A Claude Code kit with two roles that grade opposite halves of the same round. The MENTOR coaches how each request was framed before the work runs, then grades the ask — framing, leverage, learnings applied. The CRITIC judges, in fresh context and its own words, whether the thing you ASKED for actually happened, then grades the delivery — did it happen, was there evidence, did it stay in scope. Both write to one append-only ledger per repo, and the SCORECARD renders it as a published board where the human prompter and the model are scored side by side.

by Andre Paat

Post-merge verification tail: diffs merged main against the pre-merge state to catch junk files leaked by squash merges and unintended path changes, checks configurable grep-based business invariants (attribution parameters, forbidden patterns), auto-opens a cleanup PR for leaked junk, and guides an adjacent-behavior regression spot-check. Complements pre-merge review gates.

by Not Rest Inc.

A cross-runtime working-session harness for Codex and Claude: establishment, continuity, evidence-led planning, deterministic health checks, and thirty-two composable skills.

by YK

Long-horizon task management for Claude Code — intake alignment, automated orchestration, independent verification

by Sleepy Panda SRL

An operating framework for agentic engineering: a small always-loaded kernel, doctrine for the loop, autonomy tiers, verification and memory, and the Workspace Definition your team's own policy layer is validated against.

by yukineko

Config-driven pre-commit static audit for Claude Code: a Stop hook that runs generic checks plus your project rules (from a TOML file) over the pending diff — shelling out to git and optional linters — and blocks the stop, feeding findings back to the agent until they're clean. Subscription-native (one hook + bundled Rust binary, no API key). v0.1.16: the shared in-tree `<audit_dir>/.audit-skip` marker is gone — it was the fifth mirror of the one-shot bypass the four Stop gates dropped in the same change. It carried no attribution and lived in the working tree, so whichever invocation ran next spent it: a bypass one session armed was routinely consumed by a different session's commit, or by a human's terminal `git commit`. The replacement is `precommit-audit skip --reason "..."`, scoped to the issuing session via CLAUDE_CODE_SESSION_ID, reason-required, and recorded at both issue and consumption. An invocation with no session id consumes nothing at all.

by Sergey Vasilevskiy

Verifies signatures on the instruction files Claude Code loads, meaning CLAUDE.md, skills, and agent definitions, before they reach model context. It re-checks a skill bundle before the Skill tool runs it. Sigstore keyless, offline verification against a pinned trust root, no telemetry.

by Anna Rosen

Research-coding workflow for computational science, built with a JAX/Python astrophysics group and domain-agnostic in its contracts. The researcher owns the science: scientific assumptions are proposed and approved before they become load-bearing, experiments carry a derived prediction, and validation is staged (physics and analytic checks in development, published-result or reference-code comparison before release). 43 skills: model development and ideation, literature, scoping, building correctly, equation-critical sources, verification and numerical validation, Bayesian inference, code and figure review, performance, records and provenance, MyST documentation, reproducibility and release. Plus an equation-verifier agent, six slash commands, twelve self-limiting hooks (Stop gates warn by default; RWF_STRICTNESS=standard blocks), and per-domain lenses for reference-parity work. The same skill files load in Codex via a symlink.

by Skylar

Verification-first workflows for finding real bugs, closing inconsistencies, and making measured improvements. Prove-it-or-don't-report-it: every finding carries an evidence tier and a proof artifact, fixes ship a failing-then-passing regression test fixed at root cause, and inconsistencies are closed with a mechanical enforcement. v2 adds test-suite validation (flaky + mutation testing), characterization safety nets, root-cause sibling sweeps, regression bisection, and a regression guard.

by Skylar

Verification-first workflows for finding real bugs, closing inconsistencies, and making measured improvements. Prove-it-or-don't-report-it: every finding carries an evidence tier and a proof artifact, fixes ship a failing-then-passing regression test fixed at root cause, and inconsistencies are closed with a mechanical enforcement. v2 adds test-suite validation (flaky + mutation testing), characterization safety nets, root-cause sibling sweeps, regression bisection, and a regression guard.

by Andre Paat

File-based text posting with unicode lint and read-back verification — payloads always travel via files (never inline argv/heredocs), zero-width characters are rejected before posting, and every post is fetched back and byte-compared so curly quotes and non-ASCII content provably survived. Kills the empty-comment / corrupted-PR-body failure class.

by Andrey Myasnikov

Compose backlog scope and deliver selected Task Manager work with autonomous test fixtures, strategic explanations, and explicit critical-codebase acceptance.