e59700325a
fix: review fixes — dead code, zero-test gate, safeBranch regex, worktree collision guard
...
- Remove dead worktreeExists() and unused WorktreeSpec import (#1 , #2 )
- verify-gate: require counts.total > 0 — silent runner no longer passes (#4 )
- worktree-isolation: add /g flag to safeBranch regex (#5 )
- execution-pipeline: try/catch around wm.create with prune-and-retry (#7 )
- Tests updated to output parseable '1 passed' for mock test commands
69 files, 295 tests.
2026-07-05 00:45:35 +02:00
61bc109982
feat: execution pipeline — wires 5 harness components end-to-end
...
runPipeline() chains worktree -> repo-map -> agent task -> verify-gate
-> interrupt-gate -> flat-ledger. One async function, ~90 lines.
4 integration tests: pass path, fail path, deny path, repo-map injection.
69 files, 295 tests.
2026-07-05 00:23:21 +02:00
e2301e8bd3
docs: spec for execution pipeline — wires 5 harness components end-to-end
...
One function, one file (~80 lines + ~60 lines tests).
worktree → repo-map → agent task → verify-gate → interrupt-gate → ledger.
No deploy, no retry, no new dependencies.
2026-07-05 00:20:08 +02:00
14f2e509f4
docs: map Armin tool-calling article, add fresh factory live receipt
2026-07-05 00:07:20 +02:00
0236f351ec
test: add harness tests for interrupt-gate, flat-ledger, repo-mapper
...
interrupt-gate: 5 tests (stdin defaults, requireApproval)
flat-ledger: 6 tests (write/read, lastStep, summary, clear, sessionName)
repo-mapper: 4 tests (real dir, empty path, minifiedTree, treeString)
Fix FlatLedger to accept setLedgerDir() for test isolation.
Total: 68 files, 291 tests passing.
2026-07-04 19:15:38 +02:00
d4e2d3c36b
feat: add interrupt-gate, flat-ledger, repo-mapper
...
Three core harness components from the Dario blueprint:
1. interrupt-gate: stdin TTY gate for human-in-the-loop approval
before agent actions touch live branches.
2. flat-ledger: JSONL state file outside LLM context — agent reads
current state at every turn instead of remembering it.
3. repo-mapper: auto-generates minified repo tree map for system
prompt injection before any code generation.
Blueprint: hardware interrupt, state materialization, absolute
environment mapping.
2026-07-04 17:07:48 +02:00
b654b72905
feat: add verify-gate — hard verification gate for agent tasks
...
Build-test-lint gate that short-circuits on failure.
verifyAndClean() removes git worktree when gate fails.
Exports runVerifyGate, verifyAgentTask, verifyAndClean.
Blueprint: Dario Amodei model/harness split — hard automated
evaluation gates that drop/wipe on failure.
2026-07-04 17:04:18 +02:00
69a4e1c613
docs: index RSI phase receipt and expand quarantine table
2026-07-04 16:13:30 +02:00
ff79509bc3
chore: receipt external AI links and expand memory quarantine index
...
Maps Sakana Fugu, OpenAI Daybreak, Cloudflare Temporary Accounts
as advisory-only external research. Expands README memory quarantine
section with full 7-claim table and blocked/canonical trust paths.
2026-07-04 16:00:08 +02:00
2eae8f9763
docs: complete anthropic fable5 prompting guide receipt
2026-07-04 15:54:22 +02:00
90b36c9e22
chore: quarantine A-D autonomous deploy memory and index anthropic prompting guide
2026-07-04 15:33:29 +02:00
78667dd93c
feat: port opensoft/oh-my-opencode as learn-only external project review
2026-07-04 15:26:33 +02:00
73b55e1684
docs: add fresh factory readonly receipt
2026-07-03 09:00:48 +02:00
c4843c7b54
docs: map opencode and advisory memory context
2026-07-03 08:43:15 +02:00
6cbcf75654
docs: quarantine stale phase-claims and gate stale-memory refresh
2026-07-02 13:08:58 +02:00
1fcac197df
docs: index remaining historical receipts
2026-07-02 12:41:20 +02:00
1dff2e876a
docs: map fable website video workflow
2026-07-02 10:02:00 +02:00
7a9cd58d03
docs: add live factory readonly receipt
2026-07-02 09:23:44 +02:00
a6d21dda53
docs: add agentic system layer checklist
2026-07-02 09:08:27 +02:00
73371cf8f6
docs: map agentic ai systems research
2026-07-02 08:55:01 +02:00
72f4917a55
docs: index receipts and block stale deploy memory
2026-07-02 02:23:28 +02:00
2419fca7bc
docs: map claude artifact advisory source
2026-07-02 01:48:46 +02:00
577bc11a56
docs: map digg tech source signal
2026-07-02 01:28:16 +02:00
818c2be727
docs: map kelsey platform posture
2026-07-02 01:15:25 +02:00
85a8192810
docs: map adaptive tap pair red-team research
2026-07-02 01:10:36 +02:00
7e6b624378
chore: ignore local pi runtime state
2026-07-02 00:44:37 +02:00
9e25d208bd
docs: bound kramer ai quality patterns
2026-07-02 00:42:04 +02:00
367a39d829
docs: quarantine stale deploy memory
2026-07-02 00:31:26 +02:00
d91e279386
docs: map external pattern research
2026-07-02 00:17:02 +02:00
cecb264bc9
feat: review scankii external scanner safely
2026-07-01 23:02:47 +02:00
bac3c27380
feat: review ecc external harness safely
2026-07-01 22:50:49 +02:00
f6319a149d
feat: review external skills repos safely
2026-07-01 22:41:36 +02:00
9db357f3d1
feat: add structural patch receipts
2026-07-01 22:28:46 +02:00
60d7edff00
docs: capture ai quality research posture
2026-07-01 22:24:29 +02:00
5d9fb0619e
feat: add uast runtime receipts
2026-07-01 19:06:59 +02:00
f97674ae26
feat: add research map receipts
2026-07-01 18:46:32 +02:00
93c9697dbb
feat: add external evidence receipts
2026-07-01 18:00:48 +02:00
712d94c689
feat: add falsification receipts
2026-07-01 17:25:39 +02:00
b958364338
feat: quarantine stale factory memory
2026-07-01 15:19:12 +02:00
6872cb3a6d
feat: add external project review receipts
2026-07-01 14:55:18 +02:00
208d27e2cf
feat: add learning safety receipts
2026-07-01 14:30:26 +02:00
1135ef4d7c
feat: add forgejo actions coverage receipt
2026-07-01 14:05:28 +02:00
f4714f146a
feat: add forgejo cli commands
2026-07-01 13:50:34 +02:00
41da725bc5
docs: clarify agent manager boundary
2026-07-01 13:15:10 +02:00
e59a0f0ec4
fix: fail rsi on stale container canary
2026-07-01 12:32:29 +02:00
451267d8bc
feat: track fable mythos access safely
2026-07-01 11:21:15 +02:00
7054758d0d
feat: add jailbreak assessment receipts
2026-07-01 06:11:13 +02:00
09b2483919
feat: add prompt fingerprint receipts
2026-07-01 05:23:44 +02:00
0030455324
feat: add skill health receipts
2026-07-01 05:17:20 +02:00
ac800d106f
feat: add system card fields to agent traces
2026-07-01 05:06:12 +02:00
artale