All 3 cloud models converged on real concurrency bugs in the SQLite stores:
- AppendVersion (HIGH): the seq key was `SELECT MAX(seq)+1` then INSERT in two
un-transacted statements with a NON-unique index, AND the Scan error was
swallowed (seq stayed 0 on failure). Concurrent appends could both land the
same seq, silently breaking newest-first ordering. Now: one transaction, the
Scan error is propagated, the (skill_id, seq) index is UNIQUE (the loser of a
race fails loudly), and an empty SkillID is rejected.
- MarkScheduledRun / MarkAgentScheduledRun (all 3): replaced the Get→mutate→Save
read-modify-write (lost-update window) with a single atomic UPDATE using
json_set, so a concurrent Mark/edit can't clobber it. json_set keeps the JSON
blob's NextRunAt/LastScheduledRunAt consistent with the indexed column;
RFC3339Nano matches Go's time encoding so the blob still round-trips (tested).
- Open: actually applies PRAGMA busy_timeout=5000 (the doc advertised it but it
was never set) — a contended writer waits instead of erroring SQLITE_BUSY.
- budgetStore.Add: rejects NaN/Inf secondsUsed (would irrecoverably poison the
column).
Triaged-but-kept: plaintext webhook secret (documented design, high-entropy URL
key, pre-existing); SQL()/free-form `where` helpers (no untrusted input reaches
them — defense-in-depth notes only).
Core go.sum still free of host/DB deps; contrib/store green (incl. a json_set
blob-round-trip test).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
db.Audit() satisfies audit.Storage (all 17 methods) over SQLite: one indexed
row per run (+ a JSON inputs blob), one row per log event. Filter/list/walk
queries are indexed on the columns they filter (skill_id, caller_id,
parent_run_id, started_at); WalkParentChain follows parent_run_id with a
seen-set guard; LastRunBySkills is a grouped MAX.
Test covers run start/finish round-trip (inputs map + token roll-up), log
append + ordered read, parent/child + ancestor-chain walks, caller listing,
TopLevelOnly filter, and the last-run-per-skill map.
contrib/store now backs ALL four store seams — budget + persona + skill + audit
— so a host gets turnkey durable persistence (run history, budgets, agents,
skills) with zero store code. Core go.sum still has 0 sqlite refs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
db.Skills() satisfies skill.SkillStore over SQLite, same JSON-blob + indexed
columns pattern. Versions live in their own table (each SkillVersion embeds a
full Skill snapshot as JSON), ordered newest-first by an append seq.
Test: round-trip (Tools, ExposeAsChatbotTool), visibility listing
(public/shared/private with SharedWith filtered in Go), chatbot-exposed,
newest-first versions + GetVersionByID, scheduled-due query + MarkScheduledRun.
contrib/store now covers budget + persona + skill; audit store next.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
db.Personas() satisfies persona.Storage over SQLite. Each Agent is stored as a
JSON blob with extracted indexed columns (owner_id, name, webhook_secret,
chatbot_channel_filter, schedule, next_run_at) — so the WHOLE struct round-trips
(no domain<->GORM<->DB field-loss footgun) while the lookups stay indexable.
Test proves the round-trip preserves nested + map fields (SkillPalette,
StateReactEmoji), the owner/name + webhook + chatbot-filter queries, the
scheduled-due query, and MarkAgentScheduledRun clearing the due window.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Establish the nested persistence module — the architectural reason the core
stays lean: a SEPARATE go.mod carrying modernc.org/sqlite (pure Go, no cgo), so
the SQLite driver NEVER enters the executus core go.sum. A static-binary host
(gadfly) importing only the core stays static; a host wanting turnkey
persistence imports contrib/store.
- sqlite.go: store.Open(dsn) -> *DB (one SQLite file), accessor-per-seam.
- budget_store.go: db.Budget() satisfies budget.BudgetStorage; Add() does the
7-day window rollover atomically inside a transaction (concurrent Adds can't
race the read-modify-write — the in-memory store's one weak spot).
- Conformance test: budget.NewDBBudget over the SQLite store passes the SAME
rolling-window contract as the in-memory store.
- CI: a new step builds + tests contrib/store on its own AND asserts it carries
the sqlite driver the core forbids (proof the split works). Verified: core
go.sum has 0 sqlite refs; contrib/store go.sum has it.
persona/skill/audit SQLite stores follow next (same JSON-blob + indexed-columns
pattern, sidestepping the three-layer field-loss footgun).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The skill half of the persona/skill pair, as a clean redesign (not a faithful
lift of mort's 60-method skills.Storage kitchen sink):
- skill.go/skill_version.go/validate.go/inputs.go/schedule.go moved clean; the
only host couplings severed: llms.IsTierName -> model.IsTierName, and the
chatbot DefaultChatbotInputName const localized.
- store.go: a DELIBERATELY LEAN SkillStore — skill lifecycle (CRUD + visibility)
+ versioning + scheduling ONLY. The KV/file/quota sub-stores that were fused
into mort's interface are the tools/ store seams; email/channel grants stay
host concerns.
- runnable.go: Skill.ToRunnable() lowers a skill into run.RunnableAgent (flat
tool list, no palette — composition is a host concern); DueAt() helper.
- memory.go: NewMemory() — zero-dep in-process SkillStore (visibility filters,
newest-first versions).
Tests: ToRunnable mapping, visibility (public/shared/private) listing, version
ordering + lookup. No mort dependency (go.mod tidy clean); core imports ZERO
from skill.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The headline P4 piece (clean redesign): the Agent persona noun, decoupled from
its Discord shell.
- agent.go/storage.go/builtin_loader.go moved from mort's pkg/logic/agents; the
Storage seam drops the Discord CommandBindingStorage embedding (a host
concern). The host-entangled files (commands, chatbot_provider, command-
binding dispatcher, personalization, system) stay in mort.
- runnable.go: Agent.ToRunnable() lowers a persona into run.RunnableAgent — the
bridge that lets run.Executor run a persona without importing this battery
(the inversion of agentexec.Run(*agents.Agent)).
- memory.go: NewMemory() — zero-dep in-process persona Storage (all 11 CRUD +
trigger-query methods).
Tests: ToRunnable field/phase mapping; Memory round-trip. CI invariant: core
imports ZERO from persona.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Second Tier-2 battery, plugging into run.Ports.Budget:
- budget.go: skillexec's BudgetTracker / NoOpBudget / DBBudget moved clean
(stdlib only). Check/Commit match run.Budget exactly (compile-time proof in
run.go: NoOpBudget and *DBBudget are run.Budget).
- storage.go: the BudgetStorage seam + SkillBudget domain, split out of mort's
GORM file (the GORM impl stays in mort).
- memory.go: NewMemory() — zero-dependency in-process BudgetStorage with the
7-day rolling-window rollover in Add.
Tests: per-user cap enforced, window rolls over after 7 days, NoOp always
allows. CI invariant: core imports ZERO from the budget battery.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
First Tier-2 battery, plugging into run.Ports.Audit:
- storage.go/writer.go: skillaudit's Storage interface + per-run Writer moved
clean (only utils->fmt); the Writer already matches run.RunRecorder's shape.
- sink.go: Sink adapts a Storage to run.Audit (StartRun -> a run row + a Writer
wrapped as run.RunRecorder, converting run.RunStats on Close). NewSink(nil) is
equivalent to no audit. Compile-time proofs: Sink is run.Audit, recorder is
run.RunRecorder.
- memory.go: NewMemory() — a zero-dependency, queryable in-process Storage
(retains runs + logs; all 17 read/filter/purge/walk methods) so a light host
gets run history with no setup. Mort keeps its GORM Storage; contrib/store
adds durable SQLite at P4.
End-to-end test: wire audit.NewSink(audit.NewMemory()) into the executor, run an
agent, and the run is recorded with terminal status/output and queryable by
caller. CI invariant verified: core imports ZERO from the audit battery (proper
battery direction; battery imports core, never the reverse).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Measured on the P2 review: the local Macs (m1/m5) took 26–29 min with lens
timeouts and found ZERO real bugs, while the two cloud models found every
genuine finding in 6–12 min. Drop the Macs; add glm-5.2:cloud as a third
cloud reviewer. Net: faster (~29→~12 min) and higher signal.
Models: minimax-m3:cloud, deepseek-v4-flash:cloud, glm-5.2:cloud
(ollama-cloud=3 concurrency). timeout-minutes 90→30.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Independently verified all 18 gadfly findings against the code (18-agent
fan-out). Fixed the 9 real ones; the other 9 were false-positive /
hallucinated / valid-tradeoff (no change).
High:
- F1 nil model: a Models resolver returning (ctx,nil,nil) flowed into the
agent loop and nil-panicked. Now a clean error (Run never panics). +test.
- F9 compactor data-leak: renderTranscript sent tool-call args verbatim to
the summarizer (a possibly-different provider/tier); secret-bearing tool
args (mcp_call/email_send/http_*/webhook_*) are now redacted, with a doc
note that result bodies still flow (summary needs them).
Medium/minor:
- F2 compactor error path returned the folded slice, not the original msgs
(contradicting the documented non-fatal contract) -> return msgs.
- F3 RunStats.Status only ok/error; now timeout (DeadlineExceeded) /
cancelled (Canceled) via statusFor. +test.
- F4 step-zip emitted empty-name "ghost" steps when results>calls; now pairs
min(calls,results) only.
- F5 SetIteration was never called -> RunState.Iteration always 0; the step
observer now updates it each loop.
- F6 matchPending fallback was LIFO; now FIFO (matches the per-key queue).
- F7 estimateTokens had no default arm (future Part kinds counted as 0);
unknown parts now counted conservatively.
- F8 cloud_sync silently truncated >1MiB responses -> opaque JSON error; now
a clear "response exceeded N bytes" via readCapped.
- F12 step observer captured the caller ctx; now the merged runCtx.
- F13 compaction onFire was nil (doc claimed it logged); now wired to
audit LogEvent("compaction_fired").
- F11 (no pre-dispatch hook in majordomo) documented honestly as a known
limitation; F18 UsageSink doc clarified cache tokens are subsets of input.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The capstone of the run kernel: run.Executor.Run(ctx, RunnableAgent, inv)
ties model resolution + the tool registry + majordomo's agent loop +
context compaction + run-bounding + step/audit instrumentation into one
path, with every host concern behind the nil-safe run.Ports.
- run/executor.go: New(Config{Registry, Models, Defaults, Ports, Compactor,
ContextTokens, SystemHeader}) + Run -> Result{RunID, Output, Steps, Usage,
Err}. Budget gate (pre-run), model resolve, Audit StartRun/recorder
(satisfies RunTally, stamped on inv.RunState), toolbox build, step observer
(zips tool calls/results -> emitter + recorder.OnStep/OnTool), V10
detached-MaxRuntime context with caller-cancel merged back, compaction wired
from ContextTokens×ratio, audit Close + Budget Commit on a detached cleanup
ctx. Zero Ports = a bounded in-memory run (gadfly's case).
- run/executor_test.go: hermetic end-to-end run against majordomo's fake
provider (hello-world), Budget-rejection (no model call), Audit-port wiring
(StartRun + Close with terminal status/output). All green under -race.
- examples/minimal upgraded to the real "hello, agentic world" (~15 lines:
Configure tiers -> run.New -> Run -> print). README/CLAUDE.md updated.
Remaining P2 follow-ups (incremental): wire Critic/Checkpointer/PaletteSource/
Delivery into the loop, multi-phase Pipelines, and the no-tools direct path.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
- compact/compactor.go: the per-run stateful context compactor (token-threshold
gate, fast-tier middle summarisation, fold memory) lifted from mort's
skillexec/compactor.go. Self-contained; its only dependency is a ModelResolver
func (model.ParseModelForContext satisfies it) + a token threshold.
- run/steps.go: the step-emission/instrumentation (stepEmitter, tool->kind/
summary mapping with redaction, Result.Steps accumulation) from agentexec,
repointed onto executus/tool.
Both build green. With the run-loop mechanics, RunnableAgent DTO, run.Ports,
compactor, and step instrumentation now all in place, the remaining P2 work is
the run.Executor itself (wiring these + majordomo's agent loop), which makes
executus runnable.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Triaged gadfly's P1 review (advisory). Fixed the three clearly-correct,
low-risk items; the rest were pre-existing mort behavior or theoretical:
- model/call.go: recordUsage dropped fully-cached responses (input==0 &&
output==0 early-return missed CacheRead/CacheWrite-only usage, which
Anthropic/OpenAI prompt-caching bills). Guard now also checks cache tokens.
- llmmeta/helper.go: recordLedger swallowed Storage.RecordMetaCall errors;
now logs them (slog.Warn) so a non-logging Storage impl can't silently drop
audit rows.
- model/cloud_sync.go: the ollama.com limit-cache used unbounded io.ReadAll;
wrapped both reads in io.LimitReader(1 MiB) so a misbehaving endpoint can't
exhaust memory before the 15s timeout.
Noted-not-fixed (follow-ups / pre-existing mort semantics): tier_not_allowed
ledger label on resolution failure, unknown-model usage attribution, the
cloud_sync https scheme allowlist, and several theoretical/cosmetic items.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Mirror mort's updated adversarial-review.yml: m1/m5 pulled in via the
GADFLY_ENDPOINT_M1/_M5 secrets using gadfly's "foreman" provider type
(providers m1/m5; models m1/qwen3:14b, m5/qwen3.6:35b-mlx), 2 cloud models,
3-lens suite, pinned to the gadfly :sha-6e3a83c image. Header adjusted for
executus; functional config identical to mort's tested version.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Add run/ports.go: the host seams the executor will consume, every one
nil-safe so a light host runs with the zero Ports (no persistence/audit/
budget/critic/delegation/delivery) and a heavy host wires each to a battery.
Ports mirror mort's existing interfaces so the batteries implement them
directly:
- Audit + RunRecorder (mort skillaudit.Storage/Writer): StartRun -> per-run
recorder (OnStep/OnTool/LogEvent/Close), recorder satisfies RunTally.
- Budget (mort skillexec.BudgetTracker): Check / Commit.
- Critic + CriticHandle (mort agentcritic): Monitor -> handle with
RecordStep/RecordToolStart/Steer/Deadline/Stop (the loop wiring finalizes
with the executor merge).
- Checkpointer (mort agentexec.RunCheckpointer): Save/Complete/Fail.
- PaletteSource (mort SkillInvokerForPalette + AgentInvokerForPalette):
Resolve/Invoke skill + agent delegation.
Plus host-neutral RunInfo / RunStats.
This completes the P2 inversion DESIGN; the agentexec+skillexec ->
run.Executor merge that consumes these Ports is the remaining P2 work.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Stand up the executus/run kernel foundation, decoupled from mort:
- runengine.go: the shared run-loop scaffolding (MergeCancellation,
CleanupContextTimeout, RunFinalizer/FireFinalizers, RunStateAccessor) moved
from mort. The accessor's *skillaudit.Writer dependency is inverted to a
narrow run.RunTally interface (TokenStats + ToolCallsCount) — the kernel
reads live tallies without importing the audit battery.
- submit.go: the legacy submit-capture compat tool (stdlib + majordomo/llm).
- agent.go: RunnableAgent DTO — the kernel's view of "a thing to run" (tier,
prompt, caps, palette, phases, critic config). The persona Agent and saved
Skill will LOWER into this DTO so the kernel never imports a noun battery.
This is the spine of the agentexec.Run(*agents.Agent) inversion.
run/ builds with only majordomo + executus/tool. The executor merge
(agentexec+skillexec -> run.Executor) and the nil-safe run.Ports
(Audit/Critic/Budget/Checkpointer/PaletteSource) are the next P2 block.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Lifts mort's pkg/logic/llms into executus/model, decoupled from mort:
- tiers.go: the tier resolver now reads a host-supplied config.Source under
"model.tier.<name>" with host-supplied fallbacks (Configure(cfg, defaults,
ttl)), instead of convar.Manager. Tier NAMES + specs are host config; the
resolution mechanism (cache, reasoning-suffix dialect, chain validation) is
generic. No tier names hard-coded in the harness.
- sink.go: usage/trace recording inverted off mort's llmusage/llmtrace into
UsageSink / TraceSink seams + a model-owned Span, with nil-safe context
attribution helpers (WithModel/WithTraceID/WithUsageTool/WithUsageUser).
Both sinks optional (nil = off) so a light host records nothing.
- lane decoration repointed to executus/lane; utils.Errorf -> fmt.Errorf.
- call.go keeps GenerateWith[T] (instrumented structured output) — this is the
structured-output primitive; no separate structured/ package.
- llmmeta moved over model/ (the meta-LLM helper: tier allowlist + JSON retry
+ ledger). Its tests configure a minimal tier table via TestMain.
New tests cover the inversion: config overrides fallback, tier registration,
reasoning-suffix survival, nested-tier rejection, nil-sink no-ops.
Full module: go build/vet/test -race green; core go.sum still free of
gorm/redis/discordgo/sqlite.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Same setup as mort: the published gadfly:v1 image as a specialist swarm
(m1/m5 local Macs + 2 cloud models, 3-lens suite), posting one consolidated
advisory comment. Must live on main so it triggers on PRs.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The tool registry core (registry, permission model, Invocation, gated-tool
wrapper, ssrf guard, hmac, encryption, argcoerce, helpers, rootrun,
session_tools, webhook_rate_limit) had zero mort coupling — it imports only
majordomo/llm + x/crypto/hkdf — so it moves verbatim with a package rename
(skilltools -> tool). All same-package tests came along and pass; the SSRF,
gated-wrapper, encryption and output-pattern invariants are re-anchored here.
majordomo re-enters the module graph (now pinned to the latest, incl. the
front-loaded-output fix). model/ + llmmeta + structured follow next.
Docs: CLAUDE.md now requires README/examples to stay in sync with changes in
the same commit; CI skips docs/example-only pushes via paths-ignore.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
P0 has no external deps, so go.sum doesn't exist yet and
`git diff --exit-code go.mod go.sum` errored (exit 128) on the missing
path. Use `git status --porcelain` so the check survives a not-yet-created
go.sum and still catches an untidy go.mod or a newly-created go.sum.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Batteries-included agent-harness base, extracted from mort's agent layer.
This first cut establishes the module + the zero-coupling core primitives:
- lane, dispatchguard, pendingattach, run/progress.go: moved verbatim from mort
- config: host config Source seam + env-var default (nil-safe helpers)
- deliver: output-egress seam + Discard/Stdout defaults
- identity: AdminPolicy + MemberResolver seams (nil-safe)
- fanout: programmatic N×M swarm (bounded global + per-key concurrency)
- README/CLAUDE.md with the vibe-coded banner; CI with Go gates +
the "core stays majordomo+stdlib only" invariant
Core builds with stdlib only today; majordomo enters at P1 (model/structured).
go build/vet/test -race all green.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
steve