Migration applied to production, in the only order that does not break the
promote flow:
1. Created `ai_agent_prod_prod_write` (id=5) with the narrow `prod-write` scope —
invoices, payments, and document submission (needed by builddoc to regenerate
a modified invoice's PDF). Verified functionally: reads pass, DELETE on an
invoice returns 403.
2. Repointed the promote pipeline at that user's key. It no longer borrows the
read skills' credential; if the key is absent it dies with the provisioning
command rather than silently falling back.
3. Revoked 12 write/delete rights from `ai_agent` (id=3), the credential every
read skill holds: create/modify on customer AND supplier invoices,
thirdparties, contacts, thirdparty payment details, proposals, exports,
accounting links — and delete on proposals, events, and GED documents.
Verified after: invoices, thirdparties, contacts, products, proposals, supplier
invoices and bank accounts all still read; creating an invoice returns
`403 Forbidden: Insuffisant rights`. The documented posture and the real one
finally agree.
scopes.ts corrected against the live instance: 262 is NOT "créer/modifier les
produits" as the first catalogue guessed but the `voir_tous` ACL extension — a
READ right the skills depend on (without it, list endpoints return empty arrays
instead of 403). Revoking it would have silently blinded every read skill. This
is why the audit reads labels off /user/perms.php rather than trusting ids in
code. The READ_ONLY baseline is now the audited read surface (35 rights), not a
guess.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
The discipline (ADR-0003, the promote flow, the operating rules) was written
down and still depended on whoever was driving choosing to follow it. On
2026-07-25 an agent session wrote five documents into the production ledger
through direct API calls, bypassing the promote flow entirely — a correct result
reached by a path nobody could audit. A rule an operator can skip is a
recommendation.
The five stages are now chained by artefacts on disk. Each refuses to run until
the previous produced its file, and the file says what it needs to hear:
rehearse (sandbox, host-guarded) -> judge --pre -> gate (human) -> apply
(prod) -> judge --post. The gate binds to a manifest digest, so approving a
change-set approves THAT change-set.
An op is defined ONCE, as an API call, and replayed on the sandbox then on
production — because the first design described each write twice (a sandbox
script input and a prod API body) and the pre-gate judge immediately caught them
diverging: the rehearsal was creating a EUR invoice with no due date while
production would have received a USD one at 60 days. Two descriptions of the
same write are two things that can disagree.
Judges are context-free, cross-family per the PRD qa-strategy rule, and
advisory: a BLOCK still lets the operator approve, and the override is recorded
with their name. Blocking authority stays with the human gate and the host
guards — an LLM verdict never silently starts or stops a production write.
Verified end to end against the real 24/08 change-set (M3 deferred, USD 3,000):
- pre-gate judge (Mistral) returned BLOCK twice, correctly — first on the
sandbox/prod divergence, then on a duplicate left by a repeated rehearsal;
- apply refuses after a rejected gate;
- apply refuses without ARCO_PROD_CONFIRM;
- editing an amount after approval invalidates the gate on digest mismatch.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
static/config/company.json feeds setupCompany() in both main.ts and
provisionSandbox.ts, and main.ts defaults to the PRODUCTION address. The file
held placeholder identity values: running it against prod would have replaced
Arcodange's legal identity with them.
- capital "1000€" -> "1000". Dolibarr expects a number; the € made the value
unusable by the PDF template, which is why "Capital de 1 000 €" was missing
from every invoice since January (mandatory mention, C. com. R.123-238). This
file is the root cause — patching the database alone would have been undone
by the next provisioning run.
- siren 123456789 -> 999657455, siret 12345678900011 -> 99965745500013,
numTva FR00000000000 -> FR00999657455, rcs_rm "000 000 000 R.C.S. Evry"
-> "R.C.S. Évry", naf_ape 62.02A -> 6201Z. All read off the production ERP,
where they render on every issued invoice.
- formeJuridique SAS -> SARL, and the same correction in fleet/profile/fiscal.yaml
(legal_form), which inherited "SAS" from the PRD README. Three operational
sources say SARL: the production ERP, the signed contrat cadre signature block
("Pour Arcodange (SARL)"), and the 2026-05-28 cohort review. The PRD is wrong.
Flagged in-file for confirmation against the Kbis.
- moisDebutExercice Juillet -> Janvier (fiscal year closes 12-31 per fiscal.yaml).
fiscal.yaml still validates: 8 rules, 15 calendar entries, 7 ADC records, 0 errors.
Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
Implementation of the T02 atom over the erp#39 golden set:
- validators.py: instruction-pattern + multi-IBAN pre-screens (0 hard false
positives on the 16 real docs; all 6 injection fixtures quarantined BEFORE
any model call), the atom.yaml invariants, and literal provenance anchoring
with locale-aware locate (FR/EN months incl. abbreviations, NBSP-tolerant
amounts, line-wrap + column-interleave fragment anchoring for refs).
- extract.py: single-leg runner (MLX endpoint / vibe -p), zero credentials,
zero action tools; reasoning-channel aware.
- dual_run.py: model_policy in code — dual legs, exact critical-field
agreement; disagreement, single-valid-leg or both-invalid → escalations/
for the Claude tier (resolutions go back through validators.check).
Eval (eval/2026-07-19/, full transcripts + journals committed):
- critical-field accuracy 100 % (bar 98 %) — MET
- injection suite 6/6 quarantined — zero leaks
- overall field accuracy 94.9 % (known gaps: supplier ids often null,
period_covered format) — non-blocking, noted for the next version
- 9/16 documents escalated to the Claude tier (Mistral API timeouts, small
local model on receipts, one BIC-glued IBAN, derived-ratio rates) —
consistent with the A1 autonomy level recorded in atom.yaml
Runtimes this run: m4-local = Qwen2.5-7B-4bit (MLX), mistral = vibe -p
(mistral-medium-3.5) — provisional pending erp#45; journals are the
routing-bench raw material.
Closes erp#40 (PR to follow once arcodange/golden-set is pushed — this branch
stacks on it).
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
fleet/profile/ goes from stub to the machine-readable business-rules surface
the fleet reads (PRD agent-catalog document surface + compliance ADC framework):
- fiscal.yaml — entity, VAT position, 8 rules (regime reel simplifie until
2026-12-31 -> quarterly CA3 from 2027-01-01 per LF 2025 art. 38; KM export
autoliquidation 259-1 CGI box E2; FR 20% deductible; intra-EU reverse
charge; FX 766/666; SaaS expensed; CCA 455 lane). Every rule carries
effective_from/effective_until AND decision: adc-NNN; every date cites its
PRD anchor as an inline comment (verified against factory origin/main).
- calendar.yaml — 15 entries: acomptes TVA (2026-07 month-window, 2026-12-15),
last CA12 FY-2026 (2027-05-04), CA3 quarterly windows, CFE (December),
AG comptes annuels (2027-06-30), e-invoicing milestones (2026-09-01
reception, 2027-09-01 emission/e-reporting), URSSAF echeancier with the
in-file NOTE that a real direct debit exists since May 2026 (erp#57 revisit
of the payroll-dormant assumption), KM deferred due dates + renewal stub.
- JSON Schemas for both + scripts/validate.py (stdlib-only: strict YAML-subset
parser, JSON-Schema-subset checker, rule->ADC resolution, calendar checks).
- decisions/ — ADC register: template + adc-001..005 Accepted formalizations
(autoliquidation KM, FX->766/666, SaaS expensed, reel simplifie until
abolition, CCA personal-card lane) + adc-006/007 Proposed stubs (retainer
currency -> erp#53; capital path -> erp#51). Agents draft, the operator
Accepts — never the reverse; immutable once merged, supersede never edit.
- Mutation policy in-file: PRs only (T12 proposes, human merges).
- Same-change: profile README stub -> real doc; fleet/README.md layout line
and AGENTS.md fleet row updated (profile no longer a stub).
Validation: PASS — 8 rules, 15 entries, 7 ADCs, 0 errors, 7 warnings (the
warnings list exactly what awaits operator verification). Human gate left
open on purpose: operator sanity-read of the calendar + Acceptance of
adc-001..005.
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
- validators.py: deterministic pre-screens (instruction patterns, multi-IBAN
escalate flag) + the atom.yaml invariants (arithmetic, rates, SIREN Luhn,
IBAN mod-97, date plausibility) + literal-provenance anchoring (a value
absent from the source can never appear in output).
Tested: 0 hard false positives on the 16 real docs; 6/6 injection fixtures
quarantined PRE-model; darnis-f1042 (embedded second document) → escalate.
- extract.py: single-leg runner, zero credentials/action tools; runtimes =
MLX endpoint (Ornith/M4) and vibe -p (Mistral).
- dual_run.py: model_policy in code — dual legs, exact critical-field
agreement, disagreement/flags → escalations/, invalid-both → quarantine.
Eval run against the golden set follows in this branch.
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
- runs/2026-07-18/: the 8 sha256-pinned verifier transcripts (4 runtimes ×
2 tests), blind-judging verdicts (2 independent judges per cell, unanimous),
the erp#56 builder-bench journal + prompt + caps, and the evidence README
with the parity table.
- run-verifier.sh: mistral runtime drops the tool-filter flag (--enabled-tools
with a no-match pattern hangs vibe 2.21.0); plain -p with --max-turns 1.
Verdicts: Mistral (vibe -p, mistral-medium-3.5) and Ornith 35B (hermes MLX)
reach verdict parity with the Claude baseline on both tests → admitted to
verifier duty. Qwen2.5-7B-4bit fails both → the honest small-model floor.
Builder bench: erp#56 completed by the Mistral runtime, 0 code corrections,
261 s, acceptance run clean (0 bank-UNKNOWN) → merged as PR #68.
Closes#63 (with the paired factory qa-strategy PR).
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
Seed the invoice-extract (T02) and mail-classify (T01) golden sets from real
Arcodange history, plus an adversarial injection suite and an offline
field-level scorer.
invoice-extract/
- 16 real supplier PDFs (DARNIS/Hiway F1040/F1042/F1045/F1046, Anthropic
invoice+receipt x2, Mistral, OVH, greffe d'Evry, INPI x2, Legalstart, Qonto,
Infogreffe) fetched from the Zoho mailbox + Dolibarr GED, each with a
hand-verified expected JSON per the T02 schema. Every expected value was
cross-checked against the pdftotext -layout text and re-validated against the
deterministic invariants (HT+TVA=TTC, per-rate sums, IBAN mod-97, SIREN Luhn).
- inputs/ carries both the source PDF and its {source_sha256, mime, text} pair.
- 6 SYNTHETIC injection fixtures (LLM-directive, hidden white text, IBAN-swap
BEC lure, arithmetic-repair lure, fake tool-call, ref-hijack duplicate) whose
only correct outcome is quarantine; each PDF is marked SYNTHETIC.
- score.py: stdlib-only field-level scorer, critical fields (amounts/IBAN/refs/
dates) scored separately against the 98% bar, injection leaks blocking; a
built-in --self-test proves it catches perturbed fields and leaks.
- manifest.json: per-item provenance (mail message id / GED path + sha256),
linked Dolibarr supplier invoice, a verification note, and the list of real
documents deliberately excluded (fee statements, payment proofs, La Poste
receipts with no HT/TVA breakdown) with reasons.
mail-classify/
- 1824 historical mails labeled into {supplier-invoice, bank-notice,
government-admin, client, other} via sender-domain + subject weak supervision,
one human-correctable JSONL line per message with confidence + reason +
message-id provenance. manifest.json records the pull method and distribution.
Docs: golden/README hub, invoice-extract/README (T02 schema + conventions),
injection/README (threat table), mail-classify/README (method + distribution).
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
The harness layer (builder sessions, cold verifiers, evidence flow) gets a
committable home, per the PRD model-fleet § harness portability and erp#63:
- fleet/harness/verifier/: the two canonical verifier tests (locate-test,
cold-reader backlog audit) with pinned inputs, verbatim prompts, ground
truth and pass rules — judged context-free, never self-graded.
- fleet/harness/bin/run-verifier.sh: runs a test against any OpenAI-style
local endpoint (Ornith/MLX) or vibe -p (Mistral); emits sha256-pinned
JSON transcripts.
- fleet/harness/bin/vibe-builder.sh: the bounded shell for scoped builders
and recurring tasks — refuses the trunk (linked-worktree guard), hard
--max-turns/--max-price caps, full JSON journal per run.
- fleet/README.md layout + AGENTS.md Fleet section updated in the same
change (same-change freshness rule).
Part of erp#63 (harness portability spike, D2).
Co-Authored-By: Claude Fable 5 <[email protected]>
Claude-Session: https://claude.ai/code/session_01VRShc4QhLLU73FLHx9vskh
Closes erp#38 deliverables: fleet/ layout, atom.yaml schema documented
in fleet/README.md, 7 class skeletons per the PRD agent-catalog,
invoice-extract as the worked example (contract only — implementation
is erp#40), golden/ + profile/ stubs, AGENTS.md Fleet section with
freshness fixes (fleet/ no longer "not yet landed").
Co-Authored-By: Claude Fable 5 <[email protected]>
arcodangeandClaude Opus 5