Add continuous dogfood proof infrastructure for autonomy marketing unlock

[IanFrelinger]

Summary

Implements continuous, dated, Strict-path dogfood evidence infrastructure that auto-feeds the dogfood ledger. This enables autonomy marketing unlock when scorecard thresholds are met for ~7 consecutive days with dated Strict production Ed25519 passes.

Status: PR #523 (Strict+Ed25519) merged to master (966e6bf). Strict production paths now enforce Ed25519 signatures. Remaining blocker: canary sweep script is stub until real autonomy loop wiring completes.

Changes

Infrastructure (all in this PR)

• .github/workflows/dogfood-continuous-proof.yml: Scheduled GitHub Actions job

  • Weekday cadence (Mon-Fri 06:00 UTC)
  • Manual trigger via workflow_dispatch
  • Checks for Strict Ed25519 readiness (verifies RequireEd25519Signature=true in code)
  • Runs autonomy loop sweep on canary objective: rgb-hex-parse (6-digit hex parser)
  • Auto-appends dated pass/fail row to docs/dogfood-ledger.md
  • Auto-commits ledger updates with structured metadata
  • Uploads sweep artifacts (logs, proposals, campaign data)

• scripts/dogfood-continuous-proof.sh: Helper script

  • run-canary-sweep: Execute autonomy loop on rgb-hex-parse (stub until dependencies ready)
  • append-ledger: Append formatted row to ledger table
  • Designed for CI non-interactive execution
  • Documents intended flow: seed objective → Strict verify → certify → admit → capture outcome

Documentation

• docs/dogfood-ledger.md: Dated evidence table (aligned with PR Add dogfood ledger artifact for shippable-demo evidence tracking #520 format)

  • Table format: Date | Demo | Pass/Fail | Gap | Owner | Repro
  • First row: infrastructure milestone (this PR)
  • Auto-appended by CI workflow on each scheduled run
  • Documents all gaps honestly (canary sweep stub, lim-9 status TBD)

• docs/dogfood-scorecard.md: Marketing unlock thresholds (proposed defaults)

  • Last-N green: ≥80% pass rate over last 10 scheduled runs (~2 weeks)
  • Mean time-to-admit: ≤15 minutes (single canary objective)
  • Strict rejection rate: ≤70% (i.e., ≥30% proposals reach Certified state)
  • Consecutive days hold: ≥7 consecutive days meeting all thresholds
  • Self-apply bar: Fixture E2E is floor; real Ashlar hygiene PR via loop is framework proof
  • Marketing HOLD until: thresholds met + dated Strict E2E passes in ledger

• docs/DocsIndex.md: Add ledger and scorecard references

  • Positioned in "Trust loop / certification" section after certification-evidence.md
  • Links workflow path for continuous proof

Proposed Thresholds (for PM/Marketing Review)

Threshold	Proposed Default	Rationale
Last-N green %	N=10, ≥80%	Allows occasional transient failures while ensuring fundamental stability
Mean time-to-admit	≤15 minutes	Proves loop converges efficiently for small parser; headroom for larger objectives
Strict rejection rate	≤70%	Healthy rejection (witness catching defects); 30% certification proves viability
Consecutive days hold	≥7 days	Proves stability over time (~5 scheduled runs on weekday-only); catches regressions

Note: These are proposals. Adjust based on campaign 1-4 data, model selection, and risk tolerance.

Dependencies & Gaps

✅ Strict+Ed25519 (PR #523) — NOW ON MASTER

• Status: Merged (merge commit 966e6bf)
• What it does: Adds RequireEd25519Signature=true to CertificationVerifyOptions.Strict
• Impact: Strict production paths now enforce Ed25519 signatures; limitation 7 substantially closed
• Workflow readiness: Workflow will detect the Ed25519 requirement and attempt live Strict sweeps

Implementation Stub (Remaining Blocker)

The script (scripts/dogfood-continuous-proof.sh) currently contains a stub implementation that:

• Documents the intended flow (seed → certify → admit)
• Returns failure (marks as GAP) until real implementation ready
• Real implementation needs:
  • Autonomy loop host wiring (FirstFlight project or equivalent CLI command)
  • Docker available for sandbox sessions
  • Integration with existing spikes/autonomy-first-flight patterns
  • lim-9 resolution (status TBD)

This is intentional: Workflow is designed to be honest about readiness. No fake passes, no bypasses.

Design Principles

1. Fail closed: Never claim Strict production is ready until Ed25519 enforcement verified in code (NOW MET)
2. Honest gaps: GAP rows for all limitations (stub implementation, lim-9 status)
3. No weakened cert-gate: This PR does not touch or bypass any existing certification requirements
4. Coordination with merged PRs: Ledger format aligned with PR Add dogfood ledger artifact for shippable-demo evidence tracking #520; Strict+Ed25519 from PR Close Ed25519 signature-strip attack on Strict verification paths #523 now on master
5. Self-apply bar: Fixture E2E proves mechanics; real Ashlar hygiene PR proves framework

Testing

• Docs-only + workflow + script infrastructure (no product/Forge code changes)
• Workflow designed to detect Strict Ed25519 readiness via code inspection
• Script includes structured ledger append logic
• Ledger format matches PR Add dogfood ledger artifact for shippable-demo evidence tracking #520 table structure
• Updated to reflect PR Close Ed25519 signature-strip attack on Strict verification paths #523 merged status
• Does NOT claim autonomy marketing is unblocked (thresholds not yet met; stub implementation)

Testing strategy (blast radius)

• Not applicable — infrastructure only (workflow + docs + helper script)
• Workflow will run on schedule (Mon-Fri 06:00 UTC) after merge
• First scheduled run will record a GAP row (stub implementation)
• After stub is replaced with real autonomy loop wiring, workflow will record real pass/fail outcomes

Checklist

• Documentation created (dogfood-ledger.md, dogfood-scorecard.md)
• Documentation indexed (DocsIndex.md)
• Workflow follows existing GitHub Actions patterns (setup-dotnet, artifact upload, job summary)
• No breaking changes
• No new TODO or NotImplementedException (stub is explicitly documented as such)
• Honest about remaining dependencies (stub implementation, lim-9) and gaps
• Updated after PR Close Ed25519 signature-strip attack on Strict verification paths #523 merged to master

Release

• Not a versioned release — skip

Coordination

• PR Add dogfood ledger artifact for shippable-demo evidence tracking #520 (dogfood-ledger-artifact): Merged; defines ledger format; this PR aligns with that format
• PR Close Ed25519 signature-strip attack on Strict verification paths #523 (Strict+Ed25519): ✅ MERGED to master (966e6bf); workflow now detects Ed25519 requirement
• PR docs: ledger entry for #523 Strict+Ed25519 lim-7/8 close #532 (lim-7 close ledger follow-up): Separate PR; not touched by this update
• Project Manager: Owns the merge queue; this PR ready for review/merge after these updates

Next Steps (Post-Merge)

1. Replace stub: Implement real autonomy loop integration in scripts/dogfood-continuous-proof.sh
2. Collect data: Let scheduled runs populate ledger with pass/fail outcomes
3. Review thresholds: After ~2 weeks of data, PM/Marketing review proposed thresholds
4. Unlock marketing: When thresholds met for 7 consecutive days AND dated Strict passes exist, marketing can proceed

Files Changed

• .github/workflows/dogfood-continuous-proof.yml (new)
• scripts/dogfood-continuous-proof.sh (new)
• docs/dogfood-ledger.md (new, updated after master merge)
• docs/dogfood-scorecard.md (new)
• docs/DocsIndex.md (modified, updated after master merge)

Total: 5 files, ~480 insertions (updated after merging master)

---

For PM/CEO Review:

• Proposed thresholds documented in docs/dogfood-scorecard.md
• Workflow path: .github/workflows/dogfood-continuous-proof.yml
• Ledger will auto-populate on weekdays at 06:00 UTC
• ✅ PR Close Ed25519 signature-strip attack on Strict verification paths #523 merged: Strict+Ed25519 now on master
• ⚠️ Canary sweep still stub: awaits real autonomy loop wiring
• Marketing HOLD until scorecard criteria met (see scorecard doc for full unlock criteria)
• This PR implements infrastructure; does not claim autonomy is production-ready