integrate: cumulative x.klickd supply-chain (PRs #115-#120)#121
Merged
Conversation
Consolidate the x.klickd skill-pack build process into a single NON-NORMATIVE spec + operator quickstart. No release, no automation claim beyond the per-stage tool/manual/planned labels. - docs/rfcs/chimera/SUPPLY_CHAIN.md: 18-stage pipeline, version lineage, rollback/deprecation/approval-revocation, determinism + reproducibility, anti-mirage protocol, audit-trail index. Explicit claim boundary (no universal standard, no automatic GDPR/EU-AI-Act compliance, no proven benchmark superiority; 70-80% is a design target, not a guarantee). - docs/rfcs/chimera/packs/QUICKSTART.md: human-or-agent build/audit loop, multi-agent role split, shipped verification commands, truth-boundary warning (artifact_loaded + sha256_matches_manifest). - Pointer links added in RFC-009 §12, chimera/README §6, packs/README §7. Docs-only. No schema/SDK/package/CI change. No existing claim altered. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…ecord First real automation stage of the x.klickd supply-chain protocol. Adds a stdlib-only, offline generator that collects the 42 verifiable v4.1 candidate skill packs (+ manifest), enforces the loaded+sha256_matches_manifest gate, and writes two re-checkable artefacts: - .internal-skills/supply-chain/audit/audit_trail_index.json - .internal-skills/supply-chain/audit/determinism_record.json deterministic_run_id is derived only from inputs (timestamps quarantined in a non_deterministic_zone, excluded from every hash), so identical inputs yield an identical id across runs and hosts. A `check` subcommand verifies on-disk artefacts are in sync and exits non-zero on drift or on banned-claim/secret content. validation_results is left empty by design: the generator records but does not run the validation commands, so it asserts no outcomes it did not observe (anti-mirage). Only stages labelled `tool` are automated; everything else stays `planned` / `partial` / `manual` per the stage_automation map. Not a v4.1 GA release. No publish/deploy/merge/tag. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…ndidates Add the logical-diff stage of the x.klickd skill/pack supply chain: a deterministic, offline CLI that compares a previous candidate (--before) against a new one (--after) and classifies governance/guardrail/memory/ evidence/claim/public-boundary changes rather than raw JSON lines. Hard-fails (exit 1) on guardrail lowering, claim-boundary violations, or public/private-boundary violations; exit 0 when no blocking finding. The deterministic_diff_id is a sha256 over input hashes + sorted findings, clock- and host-independent. Stdlib only, no network. Adds 10 before/after fixtures and 17 tests covering each blocking and non-blocking class plus determinism and CLI exit codes. Operator README under .internal-skills/supply-chain/diff/. No claim of full end-to-end automation; downstream stages remain planned. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Implements pipeline steps 11 (license check) and 12 (source freshness) as a real, stdlib-only, offline CLI that triages skill/candidate source manifests. Classifies licenses (allowed/review/blocked/unknown) and freshness (fresh/review/stale/missing_date) for internal review, verifies local-file hashes, flags missing/non-https URLs, and emits a deterministic JSON report. Exits non-zero on blocking findings. Internal triage only: no legal advice, no compliance claim, no network I/O, no candidate generation. Independent of PRs #115/#116/#117. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
…blic) Internal documentation only, ahead of internal production. Adds docs/internal/INTERNAL_SKILL_V4_2_MAPPING.md + README capturing the validated v4.2 mapping corrections: - governance_system detailed and symmetric with memory_system - supply_chain renamed conceptually to skill_lifecycle (no completeness claim) - output_contract wired to context_graph via graph_bindings - harmonised competency/domain naming (primary/secondary_domain_competencies, domain_risk_profile, domain_output_requirements, competency_core) - explicit interactions layer + canonical end-to-end flow Non-normative, no release. Public .klickd stays v4.0.0 GA; the 42 x.klickd artefacts remain v4.1 candidates. No tag/DOI/package/schema/SDK change, no public v4.2 claim, no artefact modified. Internal track name kept out of public surfaces. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
Add scripts/generate_supply_chain_threat_model.py: an offline, stdlib-only, deterministic static analyser for x.klickd candidate manifests. Classifies declared threats across 11 categories (authority_escalation, human_veto_bypass, tool_boundary_violation, memory_poisoning, private_public_leak, evidence_weakening, unsourced_claim, unsafe_external_action, irreversible_action, compliance_overclaim, stale_or_unlicensed_source_dependency), emits required mitigations, and blocks (exit 1) on unmitigated high/critical findings. Adds 7 candidate fixtures, a 14-case pytest suite, and an internal/draft doc. NON-NORMATIVE: not a security certification, no GDPR/EU AI Act compliance claim, no benchmark/universal-standard claim, no release. Public artefacts remain v4.1. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
… gate (v0.1) Adds the next two tool-backed supply-chain stages on top of the cumulative integration branch, both NON-NORMATIVE and internal-only: - scripts/generate_supply_chain_candidate.py: config-only build_request -> candidate skill in the internal v4.2 target shape. Deterministic ids derived only from input bytes; missing domain info -> requires_human_premium_pass (never hallucinated); sources only from build_request/source_manifest. - scripts/run_supply_chain_promotion_gate.py: orchestrates threat-model, source/license, logical-diff, candidate-shape, and forbidden-claim / public-private boundary tripwires. Classifies ACCEPT / ACCEPT_WITH_REVIEW / BLOCK (exit 0/0/1, 2 usage). Reports — never runs — the premium pass; not_run checks recorded honestly. Tests: 39 new (deterministic repeatability, anti-mirage premium-pass, blocked candidate, clean accept, forbidden-claim/leak/over-claim blocks). Full suite 283 passed. Example candidate + gate report checked in. ACTION_LOG added; README integration index updated (stages moved planned -> tool-backed with literal scope notes). No release/tag/DOI/publish/deploy. No merge to main. Public stays v4.1. Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
7 tasks
feat(supply-chain): candidate generator + combined promotion gate (v0.1)
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Cumulative integration branch bringing the six independent supply-chain PRs together on one branch for review. Draft — not for merge to
main. No release, publish, tag, DOI, external action, or approval is requested or performed.Integrates (all base
main, all merged here with--no-ff, zero conflicts — file sets were disjoint):docs/rfcs/chimera/SUPPLY_CHAIN.md,packs/QUICKSTART.md) — non-normative, anti-mirage, lifecycle, rollback/deprecation, stagedtool/manual/plannedlabels.scripts/generate_supply_chain_audit.py,.internal-skills/supply-chain/audit/).scripts/generate_supply_chain_diff.py,.internal-skills/supply-chain/diff/).scripts/check_supply_chain_sources.py,.internal-skills/supply-chain/source-check/).docs/internal/) — internal target only; public artefacts remain v4.1, no public v4.2 claim.scripts/generate_supply_chain_threat_model.py,docs/supply-chain/THREAT_MODEL_GENERATOR.md).Added on this branch
.internal-skills/supply-chain/README.md— integration index mapping the four real, tool-backed stages (audit/determinism, logical diff, source/license, threat model) vs planned stages (candidate generation, promotion gate, full PII/secrets scanner, runtime enforcement).MASTER_BRIEF.md— in-repo protocol doc so future agents are not dependent on a missing workspace path. Restates anti-mirage rules, public v4.1 vs internal v4.2 boundary, no-claims / no-external-action rules, and the loaded-skill gate (artifact_loadedandsha256_matches_manifest).Real vs planned stages
Real (shipped + tested): audit/determinism · logical diff · source/license · threat model.
Planned (specified, not built): candidate generation (runner) · promotion gate · full PII/secrets scanner · runtime enforcement.
Anti-mirage / boundary
artifact_loaded = trueandsha256_matches_manifest = true. Stubs/catalog entries are not loaded skills..klickdv4.0.0 GA,x.klickdartefacts as v4.1 candidates. v4.2 is internal-only.Davincc77/klickd-aiuntouched.Test plan
pytest tests/test_supply_chain_audit.py tests/test_supply_chain_diff.py tests/test_supply_chain_sources.py tests/test_supply_chain_threat_model.py→ 63 passed.tests/: 244 passed (was 181 onmain; +63 new).python scripts/verify_xklickd_skill_packs.py→ OK (42 packs verified, hash-match manifest).python scripts/validate_v4_1_candidate_mapping.py→ OK (49 rows, 42 artefacts).python scripts/validate_v4_schemas.py→ all strict-schema validations passed.packages//benchmarks//examples//integrations/test modules fail collection under rootpytestdue to pre-existing import-path issues unrelated to this work; scope totests/for a clean signal.Next step
Not public release. Next: build the runner candidate generator + promotion gate (the two highest-leverage planned stages).
🤖 Generated by Computer