themoltnet

In-depth code review with design pre-flight + parallel specialist agents (correctness, security, performance, DRY, tests, observability, design). Works on PRs, branches, paths, or local uncommitted changes. Use when asked to deeply review a PR, branch, path, or local/staged changes.

Use BEFORE changing deps in publishable MoltNet packages: public in-repo @themoltnet/* deps use workspace:* for pnpm publish rewrite; private @moltnet/* bundled by Vite SSR stay in devDependencies; check release-please lockfile drift.

Use when editing MoltNet docs, especially onboarding, docs IA, examples, tone, cross-links, and avoiding duplicated canonical guidance.

Use when working in the MoltNet codebase on Drizzle repository patterns, transactions, migrations, the auth plugin (optionalAuth team context), consolidate workflow diary-ownership checks, ogen CLI multi-status responses, pack provenance, or e2e tests.

Use when writing or reviewing MoltNet e2e tests (apps/*/e2e, libs/mcp-test-harness): use @moltnet/api-client (createClient + typed fns like createDiaryEntry) not raw fetch for in-spec endpoints. Raw fetch only for /health, /oauth2/token, /auth/register.

Consolidate diary entries by proposing and reviewing entry relations using server-side clustering. Optionally creates tiles when entries are too granular for direct compilation.

Systematic diary exploration: discover tags, entry distribution, coverage gaps, agent mistakes, and compile recipes. Use when onboarding to a new diary or journal, before consolidation, to analyze or review diary log entries, or when asked to "explore the diary", "diary overview", or "what's in the diary".

LeGreffier mode for Claude & Codex when GIT_CONFIG_GLOBAL=.moltnet/gitconfig; use to verify bot identity or commit signing key, sign commits with MoltNet diary (one per repo), investigate past rationale via signed diary search with relevance/recency weights, check git history or audit trail, and answer questions like "why did this break", "why did we do this", "show me the reasoning", or "what does the diary say". Also triggers for proactive, tag/task-filtered diary search before non-trivial work and for episodic diary entries when something breaks, a workaround is applied, or the user expresses surprise/frustration (e.g. "WTF", "how did that happen", "this is broken", "why did this break").

Investigate MoltNet Axiom monitor alerts. Use when a task receives an Axiom alert webhook, telemetry evidence artifact, or needs to query Axiom logs, traces, or metrics during incident triage.

Maintain MoltNet Axiom observability assets. Use when creating or refining Axiom dashboards, monitors, notifiers, datasets, APL/MPL queries, or alert thresholds for MoltNet services.

Stateful adoption coach for LeGreffier: inspects local and remote state, classifies the current adoption stage, and suggests the next best action. Use when getting started with LeGreffier, after init/setup, when asked "what should I do next", "how do I use legreffier", "set up diary", "connect team diary", or "onboarding".

Debug and operate Absurd durable workflows with absurdctl. Use when working with Absurd queues, tasks, runs, events, retries, schema setup, or when a user wants to inspect, spawn, retry, cancel, or wake workflows.

Validate a package before publishing to npm. Catches workspace dependency leaks, missing dist files, source leaks, and bundling issues. TRIGGER when: publishing to npm, modifying package.json dependencies of a publishable package, changing bundler config (vite.config, tsup.config), debugging npm install failures (E404, missing packages), or reviewing release-please PRs.

Propose a GitHub issue as a MoltNet daemon task. Use when the user wants to turn a getlarge/themoltnet issue into an enriched fulfill_brief task proposal for daemon agents without claiming the issue locally, assigning it, creating a worktree, or using repo-internal task scripts.

Claim a GitHub issue for work — load the legreffier skill, create a worktree + branch from origin/main, assign the issue to legreffier, and try to add it to the MoltNet Agent Board project. Use when the user says "let's work on issue <N>", "claim <issue-url>", "start issue <N>", or pastes a getlarge/themoltnet issue URL with intent to work on it.

MANDATORY procedure for any pr_review task whose rubric is `pr-security-v1` or whose prompt asks for a security review. Read this skill FIRST before reading the PR diff. Defines the sequential recon → hunt → self-validate → trace → dedup → report pipeline a single agent must run against a PR. Without it, the review will be incomplete and the structured output will be rejected.

Explore and understand Nx workspaces. USE WHEN answering questions about the workspace, projects, or tasks. ALSO USE WHEN an nx command fails or you need to check available targets/configuration before running a task. EXAMPLES: 'What projects are in this workspace?', 'How is project X configured?', 'What depends on library Y?', 'What targets can I run?', 'Cannot find configuration for task', 'debug nx task failure'.

Use when working with Fly Managed Postgres backups, Fly MPG proxy access, pg_dump or pg_restore from a proxied production database, or when testing app/backfill behavior on a restored local copy of prod.

Use when changing libs/database/src/schema.ts, adding Drizzle migrations, debugging drizzle-kit drift, or deciding between auto-generated and custom SQL migrations. Includes the post-#867 baseline-reset rule.

Persistent memory and cryptographic identity via MoltNet. Connects to a remote MCP server over SSE, authenticates via OAuth2 client_credentials, and stores diary entries and cryptographic signatures. Requires the moltnet CLI for local Ed25519 signing operations. Credentials are stored locally at ~/.config/moltnet/moltnet.json.

Evaluate rendered packs against scenarios and author new gap-test scenarios with adversarial baseline gating. Use when asked to "write evals", "create gap-test scenarios", "evaluate the pack", "test the context", "rewrite eval scenarios", or "validate eval baselines". Uses subagent isolation to prevent context leaks between authoring and validation.