agentic-coding-reference

TDD cycle process and design-check decision tree for feature implementation. Load when implementing features using test-driven development.

Install or update the claude-pod tooling — the claude-pod command, the pod Dockerfile, the default config, the IDE-oracle preflight, and the egress filter pair — from this repo's tools/claude-pod/ into ~/.local/bin and ~/.config/claude-pod. Thin front-end for tools/claude-pod/install.sh: run its check mode to show drift, apply only on the user's approval. Use when the user asks to install claude-pod, set up the container pod, or update the pod script or image definition. Examples: "install claude-pod", "update the pod tooling", "set up the claude container".

Install or update the harness-stats tooling — harness-statusline.sh, cache-report.sh, and the cache-report skill — from this repo's tools/harness-stats/ into the user's ~/.claude/ directory. Thin front-end for tools/harness-stats/install.sh: run its check mode to show drift, apply only on the user's approval. Use when the user asks to install the cache tooling, set up the statusline, update the harness-stats scripts, or reconfigure cache reporting. Examples: "set up the statusline", "install harness-stats", "update the cache tooling".

Generate a per-agent cache-efficiency report for the current Claude Code session. Use when the user asks about cache stats, cache efficiency, prompt-cache performance, token savings, or whether subagents are using cache effectively. Examples: "show cache report", "how's the cache doing", "cache stats", "are subagents efficient", "which agents are wasting cache", "session token breakdown".

Periodic multi-angle improvement review of the reference. Always fans out all five read-only research agents in parallel — maintainer tooling, documentation, runtime cost, source duplication, consumer surface — so cross-angle tension informs the synthesis. Every agent carries the challenge charter: question the design against the README's stated philosophy, recommend removals that pay for themselves. Judges findings by the resilience-first doctrine; a settled ADR decision is rebuttable with a new argument, never silently suppressed. "/review-harness challenge" drops the ADR anchoring from the agent prompts for a zero-based outside look. Adversarially verifies structural findings and delivers one prioritized report. Assessment only — never edits. Complements /audit-harness, which gates defects against the current bar. Root-only (Claude Code).

Build, test, format, and lint requirements that must pass before code review. Load when checking implementation completeness or running the quality gate.

Security review checklists, threat model, severity classification, and supply chain verification for Java/Spring Boot applications. Load when conducting security reviews.

Test quality checklist, security testing requirements, dynamic analysis, and test organization conventions for Java/Spring Boot applications. Load when conducting test reviews.

Security-review checklist, threat model, severity classification, and supply chain verification — language-agnostic, with per-section slots a stack fills in. Load when conducting security reviews.

Hold the agentic-coding-reference to a high bar after a change; three layers, one verdict. Layer 1 runs the deterministic battery (harness/verify-harness.py — its header carries the step list). Layer 2 runs the judgment-only consistency audit — six checks, agent depth delegated to /audit-agents. Layer 3 dispatches an adversarial multi-agent review of the working-tree diff. The default run scopes the judgment layers to the diff; "/audit-harness full" runs all six consistency checks across the samples. The root CLAUDE.md maintainer loop states when each tier runs. Root-only (Claude Code).

Scaffold the project-owned files a new harness consumer commits — its CLAUDE.md rules file, .claude/settings.json, scripts/layout.toml (with the channel declaration — copy by default), the docs/ brief roster, and the .gitignore block. Detects the target's stack (Go, Java Spring Boot, or the generic fallback) from its build marker. Does NOT install the runtime; that is materialize. Never overwrites a project file that already exists. Load when the user invokes `/init <project-path>`.

Regenerate the reference's architecture figures when the harness changes, in one consistent house style. Owns the figures as native .drawio sources exported to embedded-XML PNGs; the sources are the style spec. The skill holds the intentions behind the style, each figure's composition, the draw.io export command, and the README embedding convention. Load when the pipeline, agents, loops, distribution channels, or harvest flow change, or to add a new figure. Root-only (Claude Code).

Pull generic improvements from a downstream project back into the /harness source tree. Detects the source project's stack (Go, Java Spring Boot, or the generic fallback), diffs its harness runtime (.claude/, schemas, scripts) against the materialized harness (core plus the stack slice), classifies each change as harvest, skip, or ask, generalizes domain patterns on the way back, and routes language-agnostic changes to core/ and stack-specific ones to stacks/<stack>/. Load when the user invokes `/harvest <project-path>`.

Install or upgrade a project's harness by completely replacing its harness-owned runtime with the current /harness. Auto-detects the stack (Go, Java Spring Boot, or the generic fallback) from the build marker, scaffolds project-owned files via /init when they are missing, replaces the runtime, removes stale orphans, preserves genuine project extensions (asking when unsure), respects the project's declared distribution channel, verifies the installed suites, and validates with the doctor. Load when the user invokes `/materialize <project-path>`.

Cut a new harness version. Evaluates the semantic-version bump from the conventional commits since the last v* tag, proposes it with reasoning, and asks you to confirm or override (forward-only). On confirmation it runs harness/release-version.sh, which stamps harness/VERSION (+ release date), runs propagate-harness.sh (propagate + full battery), then creates the local chore(release) commit and the annotated v<VERSION> tag. It STOPS there and prints the push commands; it never pushes or publishes unasked. Run on a clean tree once the feature work is committed. Root-only (Claude Code).

Check upstream tool documentation for changes that affect docs/cross-tool-strategy.md. Fetches source URLs, compares claims against current doc, and reports drift. Use when updating the cross-tool strategy guide or checking for tool changes.

Audit the agentic configuration for consistency, coherence, and conciseness. Load when modifying agent definitions, skills, or pipeline structure, or to verify cross-tool parity.

Pipeline routing rules and handoff conditions between specialist agents: gates, recovery, and the root-applied procedures. Load when coordinating feature delivery, checking pipeline state, or determining which agent to invoke next. The writer side of the log lives in handoff-append.

PRD format conventions, boundary rules, and template references. Load when writing or reviewing product requirements.

Review process overview, feedback tag definitions, and output format. Load when conducting or processing code reviews.

Build, test, format, and lint requirements that must pass before code review. Load when checking implementation completeness or running the quality gate.

Build, test, format, and lint requirements that must pass before code review. Load when checking implementation completeness or running the quality gate.

How to write and review project documents: the language-agnostic writing standards, the five-document architecture and ownership boundaries, the document-structure model, cross-reference and maintenance rules, and the review checklist the doc-reviewer enforces. Load when authoring or reviewing any PRD, system-design doc, ADR, brief, or CLAUDE.md.

How to write and review project documents: the language-agnostic writing standards, the five-document architecture and ownership boundaries, the document-structure model, cross-reference and maintenance rules, and the review checklist the doc-reviewer enforces. Load when authoring or reviewing any PRD, system-design doc, ADR, brief, or CLAUDE.md.

How to write and review project documents: the language-agnostic writing standards, the five-document architecture and ownership boundaries, the document-structure model, cross-reference and maintenance rules, and the review checklist the doc-reviewer enforces. Load when authoring or reviewing any PRD, system-design doc, ADR, brief, or CLAUDE.md.

Check pinned tool, plugin, and dependency versions across the Go and Java Spring Boot samples — including the init skeletons and root README that restate them — plus the SHA-pinned GitHub Actions in the root CI workflow and the dated pricing override in the harness-stats accounting, against upstream stable releases. Reports drift as a table, applies approved bumps to build files, version tables, and workflow pins, and verifies each change.

Audit the project's docs/ briefs against the high bar — each document on its own (principle form, enforceability) and in combination (cross-document coherence, contradictions, brief-vs-data agreement). Runs the doctor (deterministic structural gate) first, then the advisory judgment review, and reports both. Load when you want to check the docs hold up — at onboarding, after a harness upgrade, or on request. The judgment half is advisory: it judges form, never philosophical direction.

Grade a passing change for how much human attention it deserves before merge. Load when running the change-grader after the reviewer roster approves. Holds the full grading protocol: the five facets, worst-facet aggregation, the facets-rationale-verdict order, persistence, and scope/non-goals.

Deterministic, blocking validation of the project's docs/ brief against the harness-project API: roster existence, required sections, data slots, naming conventions, channel invariants, and the reviewer roster. Load when onboarding a project, after a harness upgrade, or before starting pipeline work. Model-free, CI-runnable.

The writer contract for the handoff log: how every pipeline agent appends records to .scratch/handoff.jsonl through scripts/handoff.py. Load when producing any handoff record — dispatch-start, a substantive record, or a consultation. The routing side lives in handoff-routing.

The reader board for the handoff log: render each slice of .scratch/handoff.jsonl — header, review-convergence matrix, timeline — to the terminal via scripts/handoff.py view. Load when the user asks where the pipeline stands, for a slice status, or for a review-progress summary. The writer side lives in handoff-append; routing lives in handoff-routing.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.

Set up or upgrade this harness plugin's project-side half. Installs the engine sliver (scripts, schemas, templates) into your project and gitignores it, and refreshes the managed CLAUDE.md chapters. Run after installing the plugin and again after every plugin update.