artale
/
fable-agent
Template
1
1
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

docs: clarify attestation proof layer positioning

This commit is contained in:
artale 2026-06-21 14:28:10 +02:00
parent 85d3788c8a
commit f0fe50a253
2 changed files with 20 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
README.md
View File

@ -1,8 +1,8 @@
# Fable Agent
**Harness-agnostic, model-agnostic, self-improving agent system — loops, dynamic workflows, routines.**
**Agentic attestation infrastructure — typed receipts, safety scanners, risk gates, and guarded deploy proof for AI-generated work.**
The harness, not the model. Swap any executor, any model, any provider — the system compounds regardless.
The harness, not the model. Pi/Codex/Claude can act; fable-agent proves what happened, what verified it, what risk remains, and whether guarded deploy is allowed.
```
node dist/index.js demo
@ -54,6 +54,10 @@ node dist/index.js demo
# Run a task
node dist/index.js run "Review the auth module for security issues" --loop 5
# Prove agent work before saying done
node dist/index.js verify . --out .fable/attestations/latest.json
node dist/index.js factory verify . --out .fable/attestations/factory-latest.json
# Web search via Exa (set EXA_API_KEY in .env or export it)
export EXA_API_KEY=your_key_here
node dist/index.js exa search "self-improving agent systems"

16
docs/agentic-layer-tac.md
View File

@ -3,7 +3,7 @@
## Golden path
```text
idea → safety gate → ADW plan/build/test/review → verify-cycle receipt → guarded deploy decision
idea → safety gate → ADW plan/build/test/review → agentic attestation → guarded 8099/deploy decision
```
## Required skills
@ -15,6 +15,7 @@ idea → safety gate → ADW plan/build/test/review → verify-cycle receipt →
| ADW pipeline | Plan → Build → Test → Review → Ship | documented skill available | run on one real change |
| verifier | Prevent fake PASS | prompt rule active | every receipt includes commands/output |
| verify-cycle | Factory/RSI/skill health proof | `factory verify-cycle` live-passed on `rsi_canary` and `apple` | non-canary degraded repair proof |
| agentic-attestation | Prove any harness output before done | `fable-agent verify`, `factory verify`, schema `fable.verification.attestation.v1` | attach to Pi/OpenClaw loops |
| skill-sync | Import skills without corrupting IDs | frontmatter + explicit ID parsing tested | clean generated duplicates |
| platform-engineer | Keep MVI discipline | active principle | no dashboard/orchestrator until receipts are boring green |
@ -22,7 +23,7 @@ idea → safety gate → ADW plan/build/test/review → verify-cycle receipt →
- Class 1 Grade 6/7: skills, workflows, subagents, mental-model docs exist.
- Class 2 Grade 1: factory webhook/out-loop path exists at `8099/deploy`.
- Class 2 Grade 2: deterministic workflow proof is emerging via `verify-cycle`.
- Class 2 Grade 2: deterministic workflow proof is emerging via `verify-cycle` and `fable.verification.attestation.v1`.
- Class 3: not yet. Do not claim orchestrator maturity until ADW runs are receipt-backed.
## Acceptance criteria for next release
@ -36,8 +37,19 @@ PASS real-engineering verify
✓ factory status ok
✓ deploy route 8099/deploy present
✓ verify-cycle receipt written
✓ agentic attestation attached
```
## Pi integration
Pi/OpenClaw/Codex/Claude loops should call fable-agent as the proof layer before final completion:
```bash
fable-agent verify . --out .fable/attestations/latest.json
```
No final `done` without an attestation or explicit `UNVERIFIED` status. Current guarded deploy path remains `8099/deploy via git-proxy`; `8098 deploy-webhook` is legacy/stale for direct fable-agent decisions.
## Do not build yet
- dashboard