An experimental cross-document consistency reconciler. It finds small-looking
contradictions that make a repository’s documentation, history, metadata, and
attribution disagree. It is registered as an interop component of the
reposystem tool ecosystem, exposing a planned reconcile action.
The name combines reconnaissance for silly inconsistencies with reconciliation: inspect every relevant surface, retain evidence for each disagreement, and propose the smallest coherent resolution.
The intended scope includes:
-
project or file renames that were not propagated through READMEs, wikis, examples, links, repository metadata, and the GitHub description;
-
changelog or release claims that do not match the change, omit consequential side effects, or continue foregrounding obsolete generations;
-
contributor credit that is missing where it belongs or incorrectly expands into repository-wide authorship;
-
inconsistent language policy (including UK, US, and Oxford English), terminology, capitalisation, analogies, and conceptual explanations;
-
references to files, faults, features, versions, commands, or names that no longer exist; and
-
explanations that cannot all be true, even when a superficial text normalisation would make them look alike.
Ambiguous conflicts are findings for review, not permission to rewrite. Every finding and proposed edit must retain its sources, evidence spans, and provenance.
ForthWall is the proposed name for an optional bounded Forth rule engine
beneath the reconciler. It is not the meaning of recon in the project name,
not the whole product, and not currently an implemented capability.
The intended wall admits only declared inputs and approved document-level words. It must not expose raw memory, arbitrary filesystem access, shell execution, networking, or unrestricted writes. It must refuse automatic repair when evidence is ambiguous, inputs are stale, parsers disagree, or coordinates have drifted.
This repository currently contains project scaffolding and an ABI/FFI seam; it
does not contain a proven reconciliation engine or an operational ForthWall
VM. The machine-readable state records both the VM and actual reconcile
action as 0% complete. Design documents and deterministic execution alone are
not evidence that a proposed edit is semantically correct.
A capability may be described as delivered only when the repository contains independently reproducible evidence for it:
-
an executable specification for the VM and every document-level word;
-
planted positive and negative fixtures for every inconsistency class;
-
exact input hashes, evidence spans, proposed changes, and replayable traces;
-
idempotence, non-interference, scope-confinement, and stale-input tests;
-
attribution tests proving local credit cannot become global authorship;
-
adversarial and cross-format end-to-end fixtures; and
-
an independent verifier for every automatically applied edit.
Until those gates exist and pass, results are proposals requiring review and
automatic mutation must remain disabled. See
TEST-NEEDS and
issue #44.
The repository remains registered in reposystem as an interop-role tool
with status = extracted (a gitlink submodule), whose sole declared action is
reconcile.
The repository is scaffolded from the Rhodium Standard Repository (RSR) template, so it ships with CI/CD, machine-readable project metadata, an AI-agent gatekeeper protocol, a formally-typed ABI/FFI seam (Idris2 + Zig), container and reproducible-build scaffolding, and governance infrastructure.
git clone https://github.com/hyperpolymath/recon-silly-ation.git
cd recon-silly-ation
# See the available tasks:
just # lists all phases (build, test, validate, audit, ...)
# Build the Zig FFI bridge:
just build
# Check the repo still satisfies the RSR shape:
just validate # structure + metadata checks-
Machine-readable metadata (
.machine_readable/6a2/) —STATE,META,ECOSYSTEM,PLAYBOOK,AGENTIC,NEUROSYM,CLADE, andanchors/ANCHOR, in a2ml, so tools and agents can read the project’s state and boundaries. -
AI gatekeeper protocol —
0-AI-MANIFEST.a2mlis the universal entry point that tells an AI agent how to work in this repo before it touches anything. -
Typed ABI/FFI seam —
src/interface/Abi/(Idris2 type + layout proofs) oversrc/interface/ffi/(Zig implementation), with generated C headers. -
CI/CD — GitHub Actions for quality, security (CodeQL, Scorecard, secret scanning), multi-forge mirroring, and RSR anti-pattern enforcement.
-
Supply-chain & reproducibility — container layering (stapeln), Nix/Guix shells, SBOM, and signing hooks.
-
Governance —
GOVERNANCE.adoc,MAINTAINERS.adoc,.github/community health files, and a releaseAUDIT.adocgate.
| Path | What lives there |
|---|---|
|
Universal entry point for AI agents (read first). |
|
Project metadata, policies, contractiles, AI configs. |
|
Typed ABI (Idris2) + FFI (Zig) + generated headers. |
|
Onboarding, status, governance, practice, decisions. |
|
Task orchestration ( |
|
Workflows + community health files. |
|
Test suites and benchmarks. |
-
EXPLAINME.adoc— the engineering deep-dive: how the pieces actually work. -
AFFIRMATION.adoc— the dated, signed honesty snapshot of the repo’s true state. -
AUDIT.adoc— the release audit gate. -
.machine_readable/ai/PLACEHOLDERS.adoc— the full placeholder reference.
Released under the Mozilla Public License 2.0 unless a per-file
SPDX-License-Identifier says otherwise.