Skill Claude CodeCodex
Execute multi-stage verification pipelines (linting, type checking, test suites, schema validation) and emit structured diagnostics for agent self-repair before declaring completion.
Skill Claude CodeCodex
Mechanically verify that AI-generated tests are genuinely falsifiable by temporarily unpatching source code to prove that tests fail when the bug is present, rejecting vacuous and tautological tests.
Agent Claude Code
Use when implementing a validated plan.md: executes the plan one atomic step per turn (tests, implementation, or documentation), runs the checks and controls until green, records each step with code-loop, commits on a short-lived branch, and opens a pull request with the pr-record body. Developers supervise and step…
Agent Claude Code
Use when drafting an Architect decision record (docs/decisions/NNNN-slug.md) from a Think record or a described decision, or an escalation with a default date, without writing code. The record is always Drafted and unsigned: the Architect decides and signs. Suitable for interactive use and headless CI workflows.
Agent Claude Code
Use when searching the Forge, the catalog of certified reusable assets, for something to reuse before a custom build (discover mode), or when evaluating a proposed asset repository against the Forge certification checklist and drafting its catalog entry for a pull request (certify mode). Never invents assets and never…
Agent Claude Code
Use when a deterministic gate failed on a pull request (clean re-run, criteria verification, mutation bar, contract compatibility, security gate, control guard, preview smoke) and the exact fault must be diagnosed and fixed within the step and file scope, without ever weakening, skipping, editing, or waiving a…
Agent Claude Code
Use when watching what a release does in production and closing the loop: scan exported telemetry for anomalies and trends and correlate deploys (detect/explain), present proposed hypothesis verdicts for the Outcome review, draft and validate incident records, answer what reached production that a control should have…
Agent Claude Code
Use when turning a plain-text client meeting transcript into a normalized transcript.md and a Meeting Pack (pack.md) with an executive summary, domain terminology, and candidate business needs backed by verbatim client quotes, without modifying code. Also prepares the Product Owner's read-back list. Suitable for…
Agent Claude Code
Use when documenting the migration of an application or platform from one infrastructure to another (on-premise to cloud, cloud to cloud, datacenter to datacenter): inventory the current state from the repository and manifests, map each component to the target, register the unavoidable application changes, plan the…
Agent Claude Code
Turn an approved Think record (think.md) into an ordered, atomic implementation checklist with file scopes, acceptance criteria, test plans, and code snippets in plan.md. Suitable for interactive planning and CI automation.
Agent Claude Code
Use when drafting the release declaration (release.md) for one merged change: tier and rollback from think.md, the outcome metric from the Triage hypothesis, rollout stages, and thresholds proposed from the tier defaults, so DevOps can approve before deployment. Never deploys, never holds production credentials, and…
Agent Claude Code
Explore options, discover Forge candidates, evaluate trade-offs, assess blast radius, and produce the Think record (think.md) from a prioritized backlog issue. Suitable for interactive exploration with the Architect and Developer.
Agent Claude Code
Use when turning a candidate business need (meeting-pack need, support-ticket cluster, audit finding, Learn signal, or field note) into an evidence-linked backlog issue with system mapping, a prefilled scorecard, and a proposed decision in triage.md, without writing code. Suitable for interactive triage and headless…
Skill Claude CodeCodex
Drive the Code stage one plan step at a time with persistent progress, enforce the fix-loop limit, and guard agent branches. Use when a Code agent implements a validated plan.md, when it records a step result or failure, or before committing to check the branch is short-lived and correctly named.
Skill Claude CodeCodex
Check the health of Persistent Context and prepare the weekly Persistent Context & Controls review: lint instruction files (missing description, broad applyTo, the everything instruction, near-duplicates, dangling skill/agent/prompt references, secrets, stale files, count growth) and keep an append-only corrections…
Skill Claude CodeCodex
Classify the differences between two OpenAPI 3.x JSON documents (or two JSON Schemas, with local $ref resolved) as BREAKING or compatible, and fail on any breaking change unless an Accepted, human-signed decision record is supplied. Use when the agent loop or CI (Prove stage) must guarantee that public interfaces and…
Skill Claude CodeCodex
Fail a change that weakens a deterministic control (deleted or skipped tests, lowered thresholds, disabled workflow steps, edited control files, added suppressions, removed assertions, production credentials in agent workflows) and enforce the bug rule (the failing check is committed before the fix). Use in the Code…
Skill Claude CodeCodex
Prove that every acceptance criterion (AC- ) in a plan.md has at least one passing test and that no test mapped to a criterion failed, by reading JUnit XML or Jest/Vitest JSON reports and printing a criterion-by-criterion verdict table. Use when the agent loop or CI (Prove stage) must gate a change on Criteria Test…
Skill Claude CodeCodex
Structure, draft, validate, and track the Architect's decision records (docs/decisions/NNNN-slug.md) and escalations with a default date. Use when a Decision Scribe agent or session drafts a decision from a Think record, when CI validates numbering, supersession, signatures, or the think.md decision-record…
Skill Claude CodeCodex
Record and compute the Agentic Software Factory metrics from an append-only event log (metrics/events.jsonl): validate and append events, then print a Markdown or JSON report of flow, per-stage, AI-leverage, and outcome metrics with targets and ok/warn/unknown status. Use when a pipeline must emit a delivery event…
Skill Claude CodeCodex
Search, bind, validate, and certify entries of the Forge, the catalog of certified reusable building blocks (shared services, libraries, templates, interface contracts). Ships a seed catalog with Gatelin and Foxnox. Use when Think looks for assets to reuse before a custom build, when CI checks Forge bindings in…
Skill Claude CodeCodex
Maintain and validate the Learn hypothesis ledger (specs/learn/hypotheses.md): add Open hypotheses, list reviews and feature deletions that are due, compute a deterministic proposed verdict from the observed metric, and report the validation and deletion rates. Use when preparing the weekly Outcome review, when a…
Skill Claude CodeCodex
Author and validate incident records (docs/incidents/ - .md): detect, mitigate first, diagnose with correlated evidence, repair with a failing test first, and learn within five working days, ending in control hardening and a Triage issue. Use when drafting a postmortem from timeline notes, when CI validates an…
Skill Claude CodeCodex
Close the Learn loop: turn validated incident records, invalidated hypotheses, and high severity telemetry findings into Triage backlog issue bodies (labels needs:triage and source:learn, marker comment learn-signal) so every loss and early warning reaches the Triage review. Use at the end of an incident review, at…
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: