praxis-stack

What this is

Claude Code (and peer harness) sessions start cold: hundreds of installed skills, agents, and MCP servers; memory spanning every project on a machine; no deterministic opening move. The praxis stack is a pair of skills that fixes this — and then audits itself.

megapraxis — the session bootstrap

megapraxis runs at session start. It detects one of seven modes from the working directory's shape and history, inventories skills/agents/MCP servers, flags silently-broken ("dark") plugins, reads project memory and git state, and emits a short dashboard briefing. Its contract: every line corresponds to a truth on disk — unverified claims are dropped, never emitted; counts that weren't computed render as count-skipped, never as guesses.

metamegapraxis — the meta-auditor

Where megapraxis briefs the session, metamegapraxis briefs megapraxis. It rolls through 26 documented operational traps (each with a disk-verifiable probe), checks the bridge manifest's referential integrity, computes coverage deltas, and — when single-lens auditing isn't enough — dispatches a four-agent audit swarm whose verdicts persist to a knowledge graph. Focus and phase selectors compose in the fixed order audit → diff → swarm → propose → apply; applying changes never implicitly authorizes an expensive swarm dispatch.

The 26 operational traps

Silent-failure classes learned in production. Live roll-call is on the Live State tab. Full probe semantics: white paper §7.

Verdicts and Tarski honesty

Audit cycles terminate with an explicit verdict — converged, converged-modulo-restart, extend, extend-stable, degrade, cycle-complete-extend-deferred, or INCOMPLETE (with reason). What can't be decided is named, never absorbed: five canonical undecidables (unknown-unknowns, true-lens-independence, fixed-point-vs-plateau, state-validity-horizon, converged-vs-correct) plus observation-class buckets appear in every audit rather than being quietly folded into "all done".

Swarms and the bridge flywheel

When one lens isn't enough — self-audit above all — the stack composes an agent swarm: four reviewers in a pipeline, a coordinator synthesizing a verdict, every contribution persisted to a graph. Recurring reasoning compiles into bridges. The 1.6.0 fingerprint covers canonical skill membership, enabled state, layer assignments, orphan state, and completeness; an unchanged re-audit re-asserts the prior verdict without creating a new invocation. Novelty still pays full price.

Assistant runtime boundary

The optional graph backend can route work through ClaudeCode, CodexCode, or GrokBuild. CodexCode keeps codex exec on the Codex installation's subscription authentication. Only its unavailable-CLI fallback uses the official OpenAI CLI with a child-scoped API key; that fallback does not reconfigure the desktop app, and no secret is included in this package or site.

Onboarding short-circuits

• Codex · cert
• Grok Build · cert — Tier 0 filesystem-fallback; traps #4/#5 N/A unless graph requested
• Technical white paper — full specification

One real cycle, end to end — with receipts

This is not a mock-up. On 2026-07-17 the stack ran its full loop against itself — /megapraxis boot → /metamegapraxis audit → four-agent self-audit swarm → coordinator verdict → fixes applied → this site republished — and every stage left records. This tab replays that cycle and labels each claim by how far you can verify it from here:

• live — your browser re-checks it against this site's own artifacts, right now, below;
• graph — read from the audit knowledge graph at build time, scrubbed (ids become short hashes);
• origin — attested by grepping the origin machine's code at build time; honestly marked as not client-checkable.

The punchline is the loop eating its own cooking: the swarm's CRITICAL finding this cycle was that a fix recorded as "done, pending restart" by the previous cycle had landed in dead code — found because the dispatch tooling failed live during this very cycle's dispatch.