A source-traceable standards toolkit for defining, inspecting, implementing and governing collaborative engineering systems on GitHub.
Version 0.1.0 is a working toolkit and reviewed-draft control catalog, not a claim of exhaustive semantic consolidation or estate-wide compliance.
95 canonical draft controls; 22 parameterized templates; six applicability profiles; a read-only repository collector; deterministic configuration checks; scoped human-review attestations; time-bounded exceptions; blocking assessment gates; source ingestion, candidate extraction, structured requirement import, source/page ledgers, crosswalk generation and change-impact analysis.
The initial acquisition contains all six pinned source archives: 13,657 artifact entries, 3,742 documentation source Markdown files, and 18,019 published English page-version entries. The extraction ledger contains 150,903 candidate blocks and 605 explicitly structured source checklist requirements. These figures describe acquisition and extraction, not completion of semantic review.
| Source | Responsibility in the corpus |
|---|---|
github/docs |
Platform capabilities, constraints, procedures, reusable content, generated-data inputs and version applicability. |
github/github-well-architected |
Architecture, security, governance, collaboration, productivity and assessment practices. |
microsoft/ghqr |
Source-defined checks, scanner implementations, manual checks and evaluator limitations. |
tmcw/github-best-practices |
Context-dependent collaboration and change-management recommendations. |
jlcanovas/gh-best-practices-template |
Repository governance, documentation and community templates. |
atapas/model-repo |
Contribution, presentation, onboarding and engagement conventions. |
Exact commits and artifact digests are in sources/sources.lock.json and the acquisition receipts. See docs/source-rights.md before redistributing upstream material.
The proposed original-intent matrix and six source charters separate each source's authority from local policy adoption. Owner approval is pending.
Python 3.11 or newer is required. The supported interface is python -m ges from this repository root; a standalone installed wheel is not provided.
Note: The commands below require the complete toolkit (ges/, tests/, requirements.txt) from the full repository. For a documentation-only checkout (PR 1A), see the full repository at implementation/v0.1.0 branch.
python -m venv .venv
. .venv/bin/activate
python -m pip install -r requirements.txt
python -m unittest discover -s tests -v
python -m ges validate
python -m ges compileSupply GH_TOKEN through your authorized secret store or runner environment when permissioned observations are needed. The collector only calls read APIs. It never changes settings, creates issues, or pushes commits.
Capture omits known credential fields recursively from API responses and drops
properties whose keys contain the active request credential. Exact echoes of that
credential are redacted in response strings and decoded file content. Omission
counts remain available on partial paginated observations; alert states, permission
metadata, revisions and coverage flags are retained. A file's blob_sha identifies
the upstream blob, while its captured content may be redacted. This is bounded
response minimization, not proof that arbitrary source text, transformed secrets
or unknown provider fields are safe to publish. Keep assessment snapshots private.
python -m ges collect --repository OWNER/REPOSITORY --output .cache/snapshot.json
python -m ges audit --snapshot .cache/snapshot.json --profile profiles/solo-software.json --output .cache/assessment.json
python -m ges gate --report .cache/assessment.jsonThe profile must be reviewed for the actual target. Missing dimensions remain unknown; private or inaccessible settings are not assumed disabled or passing. The gate returns exit code 1 for mandatory unmet requirements. --require-accepted additionally blocks draft controls. --allow-exceptions permits approved exceptions only for observed FAIL/PARTIAL results; it never hides missing observations or unknown applicability.
python -m ges sync --output .cache/sources
python -m ges corpus --snapshots .cache/sources --output .cache/corpus
python -m ges ledger --snapshots .cache/sources --corpus .cache/corpusThe sync command downloads the six exact source snapshots and published English page lists. --rendered additionally retrieves all published article bodies; --rendered-limit N is an explicitly partial bounded run. Acquisition is read-only. Raw text is a private, ignored cache, not material to copy blindly into a public repository.
source-acquisition.yml produces inspectable artifacts without committing upstream material or accepting extracted policy. It has no recurring schedule.
Published body acquisition can resume from a digest-checked private cache:
python -m ges.pages --ledger .cache/corpus/published-page-ledger.json --cache .cache/rendered-docs --workers 2 --timeout 15Every indexed body is checked against its page identity, ledger digest, content digest
and size before reuse. Partial acquisition returns a nonzero exit code and retains
the full page denominator. Eight consecutive failures stop further submissions.
No credentials are sent. Acquisition does not establish pinned source identity,
semantic review, rights clearance or acceptance. Recovery reporting accepts this
cache through --rendered-directory .cache/rendered-docs.
Recovery can optionally validate independently audited, digest-bound published
assurance receipts through paired --published-assurance and
--published-assurance-policy inputs. Source-disposition permissions do not grant
certification authority. No real assurance receipt or authority grant is supplied;
missing evidence stays unknown. See receipt contract.
Optional paired --rights-acceptance and --rights-acceptance-policy recovery
inputs validate per-file, exact-use review/audit/human/distribution attestations.
No real authority grant or clearance receipt is supplied. Triage remains pending;
machine validation is not legal permission or publication. See
rights receipt contract.
The separate --publication-manifest, --publication-register,
--publication-receipts, --publication-policy, and --publication-output-root
inputs validate the exact candidate output set and actual upstream-expression
uses for public release accounting. Supply all five together. Missing inputs
leave public release clearance UNVERIFIED; legacy per-file rights receipts
cannot clear that milestone. See the actual-use contract
and completion reporting.
python -m ges render --template templates/README.md --parameters parameters.json --output .cache/rendered/README.mdAll required parameters must be supplied. Existing output files are not overwritten. JSON and YAML outputs are syntax-checked; project-specific truth, schema requirements and operational suitability require separate verification.
python -m ges impact --old old/artifacts.jsonl --new new/artifacts.jsonl --output .cache/impact.jsonChanges reopen directly referenced controls. Add --dependencies old/dependencies.jsonl
to conservatively reopen controls through resolved include/variable dependencies.
Unresolved dependencies and conditional/version rendering remain explicit review work.
controls/catalog.json is the canonical authored control definition. generated/functional-checklist.md, source-crosswalk.md and bindings.json are generated from it. templates/ contains implementation starting points; profiles/ contains explicitly draft target policies. ges/ owns acquisition, assessment and compilation behavior. tests/ exercises false-pass prevention and deterministic outputs. docs/ records architecture, limitations, decisions and acceptance criteria. evidence/ records executed work, not aspirations.
The large source corpus and page-level review queues are workflow artifacts and a separate downloadable evidence bundle; they are not duplicated into every consumer repository.
See the A3 frozen-input receipt for the reviewed capsule fingerprint, supplement assertions, and remaining source-mapping gaps.
Do not bootstrap review workers with a fresh sync: its source archives are pinned,
but its published page lists are live. Freeze the exact existing metadata first:
python -m ges.frozen_inputs freeze --sources .cache/sources --corpus .cache/corpus --output .cache/frozen-inputsThe returned capsule fingerprint must be preserved in an independently trusted, exact-head receipt. Validation and source-only hydration require that fingerprint; the partition identifies exact artifact IDs from that capsule:
python -m ges.frozen_inputs validate --capsule .cache/frozen-inputs --capsule-sha256 TRUSTED_SHA256
python -m ges.frozen_inputs hydrate --capsule .cache/frozen-inputs --capsule-sha256 TRUSTED_SHA256 --partition .cache/partition.json --output .cache/workerPartition schema: ges.review-partition.v1, with capsule_sha256 and a nonempty,
unique artifact_ids array. Hydration downloads only the assigned pinned source
archives, preserves original inventory timestamps and page-ledger bytes, and never
fetches live page lists or runs upstream code. Binary bytes remain cache evidence,
not executable inputs. Existing output directories are never overwritten.
The capsule is a bounded metadata transport, not a full project backup, an encrypted archive, source review, rights clearance or provider admission. Raw source text, rendered bodies, credentials and transcripts are not included. Keep capsules and hydrated source copies ignored; remote preservation requires separate encrypted custody and restoration evidence. The source-review policy currently authorizes Codex disposition receipts only; a cloud worker must not impersonate that identity.
The latest secret-scanning and CodeQL reconciliation reports exact reviewed relations, unsupported assertions, operational implications and remaining source work separately from table verification and policy adoption.
Completion is tracked as separate milestones: exhaustive
six-source synthesis, public release clearance for actual reused expression,
native pilot acceptance, and estate rollout. GES v0.2 requires the first three.
Estate rollout and subsequent external-standard/implementation-ancestry ingestion
for Engineering Environment Standards have separate acceptance boundaries.
Recovery's milestone_accounting reports all four evidence states and the
separate ges_v0_2 result. Its existing project_complete field still reports
the legacy all-nine-gate program. Neither field substitutes for governed
human review, owner acceptance or release authorization. See
completion reporting.
Full semantic page/file review; semantic deduplication of every source claim; rendered-content/version assurance; exact-use publication clearance; all upstream scanner predicates implemented as local deterministic checks; all required evaluator bindings and platform feature/plan adapters; live human-review and exception services; native pilot acceptance and separately authorized cross-estate rollout. docs/remaining-work.md tracks scanner-predicate coverage separately within operational completeness and defines the open acceptance gates.
The historical recovery checkpoint passed 324 local tests on 2026-10-02; the A2
branch at 84b9b2c passed 340 tests on 2026-10-06. Neither implies 95 controls implemented
as automatic checks or 150,903 candidates reviewed. Current catalog controls remain
REVIEWED_DRAFT; 593 non-adopted generated proposals are preserved separately in
controls/review_queue.json. No native policy is silently activated. See
recovery delivery and evidence/recovery-status.json.
See CONTRIBUTING.md, GOVERNANCE.md, AGENTS.md and SECURITY.md. GitHub reports this repository as public (verified 2026-10-03). Keep the toolkit public; do not interpret visibility as rights clearance or permission to publish raw caches, transcripts or private assessments. Restricted preservation material belongs in separately encrypted remote custody, not this Git history. No additional license is assigned to upstream material through this bootstrap.