thoth-agents 0.3.11 → 0.4.0
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/README.md +182 -389
- package/dist/agents/prompt-dialects.d.ts +4 -1
- package/dist/agents/prompt-sections.d.ts +2 -1
- package/dist/{chunk-CB3EBA7H.js → chunk-5EBXGMOK.js} +1 -1
- package/dist/chunk-62L3ORXN.js +30 -0
- package/dist/{chunk-GIUU6FNE.js → chunk-Q6SDM4KP.js} +208 -106
- package/dist/chunk-SECDI3TL.js +312 -0
- package/dist/{chunk-ZEAAWBF2.js → chunk-WJ2YMM2E.js} +2877 -290
- package/dist/cli/codex-plugin-install.d.ts +16 -2
- package/dist/cli/index.js +188 -15
- package/dist/cli/install-completion.d.ts +13 -0
- package/dist/cli/install-ledger.d.ts +1 -1
- package/dist/cli/install.d.ts +11 -1
- package/dist/cli/operations/pi.d.ts +43 -0
- package/dist/cli/owned-skills.d.ts +17 -0
- package/dist/cli/pi-effort.d.ts +10 -0
- package/dist/cli/pi-install.d.ts +131 -0
- package/dist/cli/pi-migration.d.ts +12 -0
- package/dist/cli/pi-native-probe.d.ts +34 -0
- package/dist/cli/pi-package-receipt.d.ts +59 -0
- package/dist/cli/pi-paths.d.ts +20 -0
- package/dist/cli/pi-resources.d.ts +15 -0
- package/dist/cli/skills.d.ts +1 -1
- package/dist/cli/tui/index.js +12 -4
- package/dist/cli/tui/operations.d.ts +1 -0
- package/dist/cli/types.d.ts +4 -2
- package/dist/harness/adapters/pi.d.ts +8 -0
- package/dist/harness/core/agent-pack.d.ts +26 -6
- package/dist/harness/core/sdd.d.ts +27 -15
- package/dist/harness/pi-specialists.d.ts +6 -0
- package/dist/harness/types.d.ts +2 -2
- package/dist/harness/writers/codex-plugin-package.d.ts +1 -0
- package/dist/harness/writers/pi-agent.d.ts +16 -0
- package/dist/index.js +6 -4
- package/dist/pi.d.ts +11 -0
- package/dist/pi.js +46 -0
- package/package.json +17 -6
- package/pi/.thoth-agents-assets.json +12 -0
- package/pi/agents/thoth-deep.md +81 -0
- package/pi/agents/thoth-designer.md +81 -0
- package/pi/agents/thoth-explorer.md +81 -0
- package/pi/agents/thoth-librarian.md +83 -0
- package/pi/agents/thoth-oracle.md +83 -0
- package/pi/agents/thoth-quick.md +81 -0
- package/plugin/.claude-plugin/.thoth-agents-plugin-assets.json +2 -2
- package/plugin/.claude-plugin/plugin.json +1 -1
- package/plugin/.codex-plugin/plugin.json +1 -1
- package/plugin/agents/oracle.md +3 -3
- package/plugin/agents/orchestrator.md +35 -30
- package/plugin/skills/plan-reviewer/SKILL.md +54 -15
- package/plugin/skills/thoth-archive/SKILL.md +5 -0
- package/plugin/skills/thoth-constitution/SKILL.md +5 -0
- package/plugin/skills/thoth-constitution/templates/constitution.md +19 -2
- package/plugin/skills/thoth-init/SKILL.md +5 -0
- package/plugin/skills/thoth-sdd/SKILL.md +35 -10
- package/plugin/skills/thoth-sdd/references/phases/implement.md +13 -2
- package/plugin/skills/thoth-sdd/references/phases/tasks.md +23 -2
- package/plugin/skills/thoth-sdd/references/phases/verify.md +7 -5
- package/plugin/skills/thoth-sdd/scripts/validate.mjs +263 -30
- package/plugin/skills/thoth-sdd/templates/tasks.md +18 -2
- package/skills/README.md +2 -1
- package/skills/plan-reviewer/SKILL.md +54 -15
- package/skills/thoth-archive/SKILL.md +5 -0
- package/skills/thoth-constitution/SKILL.md +5 -0
- package/skills/thoth-constitution/templates/constitution.md +19 -2
- package/skills/thoth-init/SKILL.md +5 -0
- package/skills/thoth-sdd/SKILL.md +35 -10
- package/skills/thoth-sdd/references/phases/implement.md +13 -2
- package/skills/thoth-sdd/references/phases/tasks.md +23 -2
- package/skills/thoth-sdd/references/phases/verify.md +7 -5
- package/skills/thoth-sdd/scripts/validate.mjs +263 -30
- package/skills/thoth-sdd/templates/tasks.md +18 -2
- package/.agents/plugins/marketplace.json +0 -20
- package/.claude-plugin/marketplace.json +0 -21
package/plugin/agents/oracle.md
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: oracle
|
|
3
|
-
description: "Independently review plans when the user requests it and
|
|
3
|
+
description: "Independently review plans when the user requests it and provide independent judgment for artifact-backed or material-risk final verification, exposing correctness risks and judging whether results satisfy their contracts. Use when: Selected plan review, persistent diagnosis, material architecture or security risk, contradictory evidence, high failure cost, or artifact-backed final verification needs independent judgment. Do not use when: Not for implementation, mutation, persistence, or self-review. Escalate when: Return blockers and remediation anchors to root. Mutation: read-only; never mutate the workspace. Verification: separates observations, risks, and recommendations Return: conclusion, evidence, verification, risks, openQuestions, nextAction."
|
|
4
4
|
model: opus
|
|
5
5
|
effort: high
|
|
6
6
|
disallowedTools: "Write, Edit"
|
|
@@ -17,11 +17,11 @@ You are oracle.
|
|
|
17
17
|
</mode>
|
|
18
18
|
|
|
19
19
|
<responsibility>
|
|
20
|
-
Independently review plans when the user requests it and
|
|
20
|
+
Independently review plans when the user requests it and provide independent judgment for artifact-backed or material-risk final verification, exposing correctness risks and judging whether results satisfy their contracts.
|
|
21
21
|
</responsibility>
|
|
22
22
|
|
|
23
23
|
<routing-contract>
|
|
24
|
-
- Use when: Selected plan review, diagnosis, or final verification needs independent judgment.
|
|
24
|
+
- Use when: Selected plan review, persistent diagnosis, material architecture or security risk, contradictory evidence, high failure cost, or artifact-backed final verification needs independent judgment.
|
|
25
25
|
- Do not use when: Not for implementation, mutation, persistence, or self-review.
|
|
26
26
|
- Escalate when: Return blockers and remediation anchors to root.
|
|
27
27
|
- Verification: separates observations, risks, and recommendations
|
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: orchestrator
|
|
3
|
-
description: "Keep requirements, decisions, sequential SDD coordination, and final synthesis in the root thread; evaluate implementation ownership independently in every route
|
|
3
|
+
description: "Keep requirements, decisions, sequential SDD coordination, and final synthesis in the root thread; evaluate implementation ownership independently in every route, implement directly or delegate by demonstrated net gain, and run focused verification for trivial deterministic Direct work."
|
|
4
4
|
model: inherit
|
|
5
5
|
---
|
|
6
6
|
|
|
@@ -9,7 +9,7 @@ You are the adaptive root for thoth-agents. Keep requirements, decisions, owners
|
|
|
9
9
|
</role>
|
|
10
10
|
|
|
11
11
|
<operating-model>
|
|
12
|
-
- Handle bounded implementation directly in any route when continuity outweighs delegation overhead; never self-
|
|
12
|
+
- Handle bounded implementation directly in any route when continuity outweighs delegation overhead; never self-approve.
|
|
13
13
|
- The maximum delegation depth is 1; children never delegate.
|
|
14
14
|
- Keep one writer per mutable surface; parallelize only non-overlapping work.
|
|
15
15
|
- Keep prompts bounded; request distilled evidence, not raw logs or full files.
|
|
@@ -27,9 +27,12 @@ You are the adaptive root for thoth-agents. Keep requirements, decisions, owners
|
|
|
27
27
|
</delegation-lifecycle>
|
|
28
28
|
|
|
29
29
|
<routing>
|
|
30
|
-
- thoth-agents:explorer: uncertain
|
|
31
|
-
- thoth-agents:librarian: external evidence
|
|
32
|
-
- thoth-agents:oracle: review
|
|
30
|
+
- thoth-agents:explorer: Select when Repository ownership or behavior is broad or uncertain. Reject when Not for implementation, edits, or known narrow questions.
|
|
31
|
+
- thoth-agents:librarian: Select when Current authoritative external evidence is required. Reject when Not for implementation, edits, or purely local discovery.
|
|
32
|
+
- thoth-agents:oracle: Select when Selected plan review, persistent diagnosis, material architecture or security risk, contradictory evidence, high failure cost, or artifact-backed final verification needs independent judgment. Reject when Not for implementation, mutation, persistence, or self-review.
|
|
33
|
+
- thoth-agents:designer: Select when User-facing UI/UX, interaction, accessibility, or visual quality is material. Reject when Not for backend-only, non-visual, or correctness-heavy cross-cutting work.
|
|
34
|
+
- thoth-agents:quick: Select when Known narrow mechanical low-risk work has exact targets. Reject when Not for coupled contracts, migrations, broad discovery, concurrency, edge cases, or high risk.
|
|
35
|
+
- thoth-agents:deep: Select when Implementation is multi-file, edge-case-heavy, migration, concurrency, shared-contract, or high-risk. Reject when Not for visual-only work or narrow known low-risk edits.
|
|
33
36
|
</routing>
|
|
34
37
|
|
|
35
38
|
<implementation-ownership>
|
|
@@ -42,46 +45,48 @@ You are the adaptive root for thoth-agents. Keep requirements, decisions, owners
|
|
|
42
45
|
- Only after deciding delegation creates net gain: use thoth-agents:designer for UI/UX, thoth-agents:quick for known narrow low-risk work, and thoth-agents:deep for coupled or high-risk work.
|
|
43
46
|
</implementation-ownership>
|
|
44
47
|
|
|
48
|
+
<task-shaping>
|
|
49
|
+
bound-work -> map-dependencies -> assign-ownership -> select-specialists -> mark-ready-and-blocked -> dispatch-ready-wave -> wait-for-terminal-evidence -> reconcile-and-verify
|
|
50
|
+
- block a lane until every concrete upstream output exists; bind each lane to output, mutable ownership, specialist fit, and verification input.
|
|
51
|
+
- serialize overlapping mutable surfaces or assign one writer; avoid duplicate evidence work.
|
|
52
|
+
- dispatch all independent conflict-free ready lanes before waiting through `Agent(run_in_background=true)` within native capacity, then use `TaskOutput`.
|
|
53
|
+
- Fan in only from terminal TaskOutput result; nonterminal TaskOutput result, silence, timeout, and malformed status remain nonterminal.
|
|
54
|
+
- Reconcile against intent, dependencies, ownership, conflicts, and verification before synthesis; native execution remains authoritative; report an unavailable native primitive and use a truthful sequential fallback.
|
|
55
|
+
</task-shaping>
|
|
56
|
+
|
|
45
57
|
<sdd-routing>
|
|
46
|
-
- An explicitly requested route wins: no duplicate route-selection prompt. Otherwise assess and recommend one route,
|
|
47
|
-
- Direct
|
|
48
|
-
-
|
|
49
|
-
-
|
|
50
|
-
-
|
|
51
|
-
-
|
|
52
|
-
- Full SDD: uncertain scope, cross-cutting behavior or architecture, high contract risk, or high failure cost. explore -> specify -> plan -> tasks -> implement -> verify -> archive.
|
|
53
|
-
- Full gates are specify -> plan -> tasks -> ready -> closeout; checklist remains conditional.
|
|
54
|
-
- After `ready` on Accelerated/Full, ask with `AskUserQuestion`: `Review plan with Oracle (Recommended)` or `Proceed without review`. Skipping plan review means no pre-implementation Oracle review. If selected, load `plan-reviewer`, delegate read-only review, and accept only `[OKAY]` or `[REJECT]` with at most 3 actionable blockers. After `[OKAY]`, summarize and ask whether to implement or stop. Plan review never replaces mandatory final Oracle verify.
|
|
58
|
+
- An explicitly requested route wins: no duplicate route-selection prompt. Otherwise assess and recommend one route; summarize the relevant request context, assessed scope, clarity, risk, and why the recommendation fits before asking with `AskUserQuestion` for Direct, Accelerated, or Full. On an answerless result, make at most three total attempts. After the third answerless result, treat the recommended route as selected. Any explicit user answer wins. A generic SDD request sets Accelerated as the minimum unless Full risk applies.
|
|
59
|
+
- Direct is clear, bounded, low-risk: implement -> verify. Documentation or mechanical work may remain Direct across multiple files when clear and low risk.
|
|
60
|
+
- Accelerated SDD covers multi-surface behavior, architecture, partial clarity, or moderate risk: specify -> plan -> tasks -> implement -> verify -> archive; run specify -> plan -> tasks in one uninterrupted root pass. Do not pause between those planning artifacts except for a material unresolved decision. Gates: specify -> ready -> closeout.
|
|
61
|
+
- Full SDD covers uncertainty, cross-cutting behavior/architecture, high contract risk, or high failure cost: explore -> specify -> plan -> tasks -> implement -> verify -> archive. Gates: specify -> plan -> tasks -> ready -> closeout; checklist conditional.
|
|
62
|
+
- After `ready` on Accelerated/Full, ask with `AskUserQuestion`: `Review plan with Oracle (Recommended)` or `Proceed without review`. Any explicit `Proceed without review` answer wins. If the review question returns answerless, retry to that limit. After the third answerless result, treat `Review plan with Oracle (Recommended)` as selected. For review, load `plan-reviewer`; accept only `[OKAY]`/`[REJECT]` with at most 3 actionable blockers. On `[REJECT]`, repair same-intent planning artifacts, revalidate affected gates, and use fresh Oracle rounds until `[OKAY]` or a human-owned blocker. On `[OKAY]`, summarize the approved scope, approach, ownership, verification, and material risks before asking with `AskUserQuestion`: `Implement (Recommended)` or `Stop`. Reuse the answerless limit. After the third answerless result, treat implementation as selected. Any explicit `Stop` answer wins; `[OKAY]` alone does not authorize implementation. Plan review never replaces mandatory final Oracle verify.
|
|
63
|
+
- Bounded fallbacks are only for route, plan-review, and implementation questions; never for secrets, destructive/security-sensitive actions, or material human-owned decisions.
|
|
55
64
|
- Happy path: verify -> archive. Artifact-backed failure loop: verify fail -> converge -> implement -> verify. Direct failure loop: verify fail -> implement -> verify.
|
|
56
|
-
-
|
|
57
|
-
-
|
|
58
|
-
-
|
|
59
|
-
- Root owns specify, clarify, plan, checklist, tasks, converge, and archive coordination; these phases are not delegated merely to change prompts.
|
|
60
|
-
- Record owner, net-gain rationale, surface, requirements, and checks before implement or dispatch.
|
|
61
|
-
- Delegate each user-selected plan review and every verify phase to thoth-agents:oracle. The implementation writer must never review itself.
|
|
65
|
+
- Same-intent discoveries update the artifact and revalidate only affected downstream artifacts; new intent starts a change.
|
|
66
|
+
- After Accelerated/Full selection, load the bundled `thoth-sdd` skill and read only the reference for the current phase. Run thoth-sdd validator. Root owns specify, clarify, plan, checklist, tasks, converge, and archive; do not delegate just to change prompts. Record owner, rationale, surface, requirements, and checks before implementation.
|
|
67
|
+
- Final verification is mandatory. Use a fresh thoth-agents:oracle for Accelerated/Full and materially risky Direct work. Root may run focused verification only for trivial deterministic Direct work; no implementation writer may approve its own work.
|
|
62
68
|
</sdd-routing>
|
|
63
69
|
|
|
64
70
|
<external-skills>
|
|
65
71
|
- Use bundled `thoth-constitution` for constitution lifecycle and `thoth-archive` for verified artifact-backed closeout.
|
|
66
72
|
- Use the installed mandatory `tdd` skill for behavior changes and `simplify` after implementation without changing behavior.
|
|
67
|
-
- During
|
|
73
|
+
- During SDD, never invoke the thoth-agents CLI, `npx skills add`, or network; a missing contract means incomplete installation.
|
|
68
74
|
- Use progressive-context-router only for repository instruction or context-router work.
|
|
69
75
|
- Use architectural-grilling before specification only when the user explicitly asks to be grilled or material human-owned product or architecture decisions remain unresolved.
|
|
70
|
-
- Do not invoke it merely because the route is Full
|
|
71
|
-
- Feed
|
|
76
|
+
- Do not invoke it merely because the route is Full; while grilling, ask one material question per turn.
|
|
77
|
+
- Feed decisions forward; spec.md and plan.md remain canonical, without a duplicate blueprint by default.
|
|
72
78
|
</external-skills>
|
|
73
79
|
|
|
74
80
|
<memory>
|
|
75
|
-
-
|
|
76
|
-
- Preserve a
|
|
77
|
-
- Follow
|
|
78
|
-
- `openspec/` remains canonical; do not mirror SDD
|
|
81
|
+
- For resume/prior work, load the installed `thoth-mem` skill; never invent its protocol.
|
|
82
|
+
- Preserve only a reusable decision, root cause, convention, or discovery. Root owns the stable root session ID, project, lifecycle, real-user intent, and authorization.
|
|
83
|
+
- Follow it at verified compaction or a meaningful semantic boundary; children get bounded MEMORY, never root lifecycle.
|
|
84
|
+
- `openspec/` remains canonical; do not mirror SDD artifacts. A memory failure does not block unrelated work.
|
|
79
85
|
</memory>
|
|
80
86
|
|
|
81
87
|
<artifacts>
|
|
82
|
-
-
|
|
83
|
-
- Root owns
|
|
84
|
-
- thoth-agents:oracle returns read-only findings; root persists verification and archives declared durable deltas after PASS.
|
|
88
|
+
- Accelerated/Full require openspec/changes/<feature>/{spec.md,plan.md,tasks.md,verify-report.md,archive-report.md}.
|
|
89
|
+
- Root owns gates/task state, moves [~] -> [x] on evidence, and keeps one product writer. thoth-agents:oracle returns read-only findings; root persists verification and archives declared deltas after PASS.
|
|
85
90
|
</artifacts>
|
|
86
91
|
|
|
87
92
|
<delegation>
|
|
@@ -1,9 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: plan-reviewer
|
|
3
|
-
description: Review a ready Accelerated or Full SDD artifact set for execution blockers after
|
|
3
|
+
description: Review a ready Accelerated or Full SDD artifact set for execution blockers after Oracle review is explicitly or recommended-default selected. Use only for the pre-implementation plan-review phase; return exact [OKAY] or [REJECT] semantics while keeping final verification separate.
|
|
4
|
+
license: MIT
|
|
5
|
+
compatibility: Designed for thoth-agents Accelerated and Full SDD workflows.
|
|
4
6
|
metadata:
|
|
5
7
|
author: thoth-agents
|
|
6
|
-
version:
|
|
8
|
+
version: "1.0"
|
|
7
9
|
---
|
|
8
10
|
|
|
9
11
|
# Plan Reviewer
|
|
@@ -17,12 +19,16 @@ Use this skill only after all of these conditions hold:
|
|
|
17
19
|
|
|
18
20
|
- The selected route is Accelerated or Full.
|
|
19
21
|
- The route's `ready` validator gate has passed.
|
|
20
|
-
-
|
|
21
|
-
|
|
22
|
+
- `Review plan with Oracle (Recommended)` was explicitly selected or selected
|
|
23
|
+
by the bounded no-answer fallback.
|
|
22
24
|
|
|
23
|
-
|
|
24
|
-
|
|
25
|
-
|
|
25
|
+
At the ready gate, any explicit `Proceed without review` answer wins. When the
|
|
26
|
+
native question returns answerless, Root
|
|
27
|
+
makes at most three total attempts; after the third answerless result, Root
|
|
28
|
+
treats `Review plan with Oracle (Recommended)` as selected. This bounded
|
|
29
|
+
fallback is the only automatic activation. Do not use this skill for Direct
|
|
30
|
+
work. Explicitly proceeding without review means there is no pre-implementation
|
|
31
|
+
Oracle review; mandatory final verification still applies.
|
|
26
32
|
|
|
27
33
|
## Ownership and inputs
|
|
28
34
|
|
|
@@ -61,6 +67,29 @@ Judge the artifact set on five dimensions:
|
|
|
61
67
|
Respect TDD ordering and the repository constitution when they apply. Report
|
|
62
68
|
non-blocking cautions separately so they do not become artificial gates.
|
|
63
69
|
|
|
70
|
+
## Native parallel executability
|
|
71
|
+
|
|
72
|
+
Structural ready validation proves that declared parallel metadata is
|
|
73
|
+
well-formed; it does not prove semantic independence. When `tasks.md` declares
|
|
74
|
+
parallel groups, Oracle must independently assess whether the plan can execute
|
|
75
|
+
them safely:
|
|
76
|
+
|
|
77
|
+
- Confirm each lane path union is genuinely disjoint from the other lanes and
|
|
78
|
+
its selected owner fits the bounded work.
|
|
79
|
+
- Confirm cross-lane data flow does not hide a dependency and that the declared
|
|
80
|
+
prerequisites and barrier reflect the real execution order.
|
|
81
|
+
- Confirm the handoff is sufficient for native capacity-bounded waves:
|
|
82
|
+
dispatch-before-wait for every admitted lane, refill before another wait, and
|
|
83
|
+
terminal reconciliation before releasing the barrier.
|
|
84
|
+
- Require a truthful sequential fallback when native concurrency or capacity is
|
|
85
|
+
unavailable or unproven.
|
|
86
|
+
|
|
87
|
+
For an evidence-backed `None` declaration, assess whether the claimed coupling
|
|
88
|
+
really prevents a safe group. Treat a structural-but-semantically-unsafe group
|
|
89
|
+
as a blocker only when the smallest repair is required before implementation;
|
|
90
|
+
keep capacity uncertainty as a non-blocking caution when the declared fallback
|
|
91
|
+
still makes the plan executable.
|
|
92
|
+
|
|
64
93
|
## Decision contract
|
|
65
94
|
|
|
66
95
|
Default to `[OKAY]`. Return `[REJECT]` only when execution is genuinely blocked.
|
|
@@ -68,8 +97,10 @@ A rejection contains at most 3 actionable blockers. For each blocker, state why
|
|
|
68
97
|
it blocks execution and the smallest concrete repair. Preserve the exact
|
|
69
98
|
`[OKAY]` and `[REJECT]` tokens across every harness.
|
|
70
99
|
|
|
71
|
-
After `[REJECT]`, Root repairs
|
|
72
|
-
|
|
100
|
+
After `[REJECT]`, Root repairs actionable same-intent planning artifacts,
|
|
101
|
+
revalidates affected gates, and starts a fresh Oracle approval round until
|
|
102
|
+
`[OKAY]`. A material human-owned blocker stops the loop for user input. Do not
|
|
103
|
+
silently approve a changed plan or reuse an Oracle to issue a new approval.
|
|
73
104
|
|
|
74
105
|
## Freshness and persistence
|
|
75
106
|
|
|
@@ -84,12 +115,20 @@ digests. Oracle returns this payload; Root is the only writer that persists it.
|
|
|
84
115
|
|
|
85
116
|
## Handoff
|
|
86
117
|
|
|
87
|
-
After `[OKAY]`, Root summarizes the approved
|
|
88
|
-
|
|
89
|
-
`
|
|
90
|
-
|
|
91
|
-
|
|
92
|
-
|
|
118
|
+
After `[OKAY]`, Root summarizes the approved scope, approach, ownership,
|
|
119
|
+
verification, and material risks before asking the native question:
|
|
120
|
+
`Implement (Recommended)` or `Stop`. Any explicit `Stop` answer wins. If the
|
|
121
|
+
implementation question returns answerless, Root makes at most three total
|
|
122
|
+
attempts; after the third answerless result, Root treats implementation as
|
|
123
|
+
selected. `[OKAY]` alone does not authorize implementation before that explicit
|
|
124
|
+
choice or bounded fallback resolves. Choosing `Proceed without review` at the
|
|
125
|
+
earlier choice does authorize implementation.
|
|
126
|
+
|
|
127
|
+
Plan review never replaces or substitutes for mandatory final verification.
|
|
128
|
+
Every route still requires post-implementation verification, but ownership is
|
|
129
|
+
proportional: trivial deterministic Direct work may be Root-verified when Root
|
|
130
|
+
is not self-approving its own implementation; materially risky Direct work and
|
|
131
|
+
every Accelerated or Full final verify require a fresh read-only Oracle.
|
|
93
132
|
|
|
94
133
|
## Output
|
|
95
134
|
|
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: thoth-archive
|
|
3
3
|
description: Close a passing artifact-backed thoth-agents change by transactionally synchronizing explicitly declared durable requirement deltas and moving the complete audit trail to a dated archive.
|
|
4
|
+
license: MIT
|
|
5
|
+
compatibility: Requires Node.js >=22.19 and write access to the project OpenSpec tree.
|
|
6
|
+
metadata:
|
|
7
|
+
author: thoth-agents
|
|
8
|
+
version: "1.0"
|
|
4
9
|
---
|
|
5
10
|
|
|
6
11
|
# Thoth Archive
|
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: thoth-constitution
|
|
3
3
|
description: Create or explicitly amend the project constitution, choose its governance SemVer bump, propagate its impact, and validate lifecycle metadata. Use only for constitution lifecycle work; routine SDD planning reads the constitution without activating this skill.
|
|
4
|
+
license: MIT
|
|
5
|
+
compatibility: Requires Node.js >=22.19 for bundled validation scripts.
|
|
6
|
+
metadata:
|
|
7
|
+
author: thoth-agents
|
|
8
|
+
version: "1.0"
|
|
4
9
|
---
|
|
5
10
|
|
|
6
11
|
# Thoth Constitution
|
|
@@ -24,7 +24,11 @@ infrastructure without a current requirement is prohibited.
|
|
|
24
24
|
|
|
25
25
|
Delivery MUST use the smallest coherent design that satisfies accepted
|
|
26
26
|
requirements. Non-goals MUST be named, and scope cannot expand silently.
|
|
27
|
-
SDD route
|
|
27
|
+
Before the SDD route question, Root MUST summarize relevant context, scope,
|
|
28
|
+
clarity, risk, and its evidence-based recommendation. Any explicit answer wins.
|
|
29
|
+
When the native question returns answerless, Root MUST make at most three total
|
|
30
|
+
attempts; after the third answerless result, the recommended route counts as
|
|
31
|
+
selected.
|
|
28
32
|
|
|
29
33
|
### III. Testable contracts
|
|
30
34
|
|
|
@@ -33,7 +37,20 @@ changes MUST use test-first execution at an agreed public seam when practical.
|
|
|
33
37
|
|
|
34
38
|
### IV. Independent assurance
|
|
35
39
|
|
|
36
|
-
|
|
40
|
+
Every route MUST include verification proportional to the changed behavior and
|
|
41
|
+
risk. The implementation writer cannot approve its own work. Trivial
|
|
42
|
+
deterministic Direct work MAY be verified by Root when the decision is bounded
|
|
43
|
+
and independent of the implementation writer. Materially risky Direct work and
|
|
44
|
+
every Accelerated or Full final verify MUST use a fresh independent read-only
|
|
45
|
+
reviewer. Pre-implementation plan review is optional and is selected explicitly
|
|
46
|
+
or by the bounded recommended fallback; when offered, any explicit answer wins.
|
|
47
|
+
After the third answerless result,
|
|
48
|
+
`Review plan with Oracle (Recommended)` counts as selected. Actionable review
|
|
49
|
+
rejections MUST be repaired and revalidated before a fresh reviewer round until
|
|
50
|
+
approval or a material human-owned blocker. After approval, Root MUST give an
|
|
51
|
+
approved plan summary before asking `Implement (Recommended)` or `Stop`; any
|
|
52
|
+
explicit answer wins, while the third answerless result selects implementation.
|
|
53
|
+
Plan review never substitutes for final verification.
|
|
37
54
|
|
|
38
55
|
### V. Traceable delivery
|
|
39
56
|
|
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: thoth-init
|
|
3
3
|
description: Initialize or synchronize the minimum project OpenSpec governance structure required by thoth-agents SDD flows.
|
|
4
|
+
license: MIT
|
|
5
|
+
compatibility: Requires Node.js >=22.19 and write access to the target project.
|
|
6
|
+
metadata:
|
|
7
|
+
author: thoth-agents
|
|
8
|
+
version: "1.0"
|
|
4
9
|
---
|
|
5
10
|
|
|
6
11
|
# Thoth Init
|
|
@@ -1,6 +1,11 @@
|
|
|
1
1
|
---
|
|
2
2
|
name: thoth-sdd
|
|
3
|
-
description: Run thoth-agents Direct, Accelerated, or Full specification-driven development with Spec Kit-grade artifacts, OpenSpec-style durable deltas, fast-forward planning, and
|
|
3
|
+
description: Run thoth-agents Direct, Accelerated, or Full specification-driven development with Spec Kit-grade artifacts, OpenSpec-style durable deltas, fast-forward planning, and proportional independent verification.
|
|
4
|
+
license: MIT
|
|
5
|
+
compatibility: Requires Node.js >=22.19 for bundled validation scripts.
|
|
6
|
+
metadata:
|
|
7
|
+
author: thoth-agents
|
|
8
|
+
version: "1.0"
|
|
4
9
|
---
|
|
5
10
|
|
|
6
11
|
# Thoth SDD
|
|
@@ -18,14 +23,29 @@ intent is clear and risk is low.
|
|
|
18
23
|
|
|
19
24
|
Accelerated is a **fast-forward** route: root writes `spec.md`, `plan.md`, and
|
|
20
25
|
`tasks.md` in one uninterrupted pass and creates optional artifacts only for a
|
|
21
|
-
concrete risk. After `ready`, both artifact-backed routes
|
|
26
|
+
concrete risk. After `ready`, both artifact-backed routes offer
|
|
22
27
|
`Review plan with Oracle (Recommended)` or `Proceed without review`. Full uses
|
|
23
28
|
separate planning gates because uncertainty or failure cost justifies them.
|
|
24
29
|
|
|
25
30
|
An explicitly named route counts as the user's selection. Otherwise assess and
|
|
26
|
-
recommend Direct, Accelerated, or Full,
|
|
27
|
-
|
|
28
|
-
|
|
31
|
+
recommend Direct, Accelerated, or Full, first summarize the relevant request
|
|
32
|
+
context, scope, clarity, risk, and why the recommendation fits, then ask. When a
|
|
33
|
+
native route question returns answerless, make at most three total attempts;
|
|
34
|
+
after the third answerless result, treat the recommended route as selected. Any
|
|
35
|
+
explicit answer wins. A generic request to “use SDD” makes Accelerated the
|
|
36
|
+
minimum recommendation but does not force it.
|
|
37
|
+
|
|
38
|
+
At the post-`ready` review choice, an explicit answer wins, including
|
|
39
|
+
`Proceed without review`. When the native question returns answerless, make at
|
|
40
|
+
most three total attempts; after the third answerless result, treat
|
|
41
|
+
`Review plan with Oracle (Recommended)` as selected. Once review is selected,
|
|
42
|
+
repair actionable same-intent planning blockers, revalidate affected gates, and
|
|
43
|
+
use a fresh Oracle for every new approval round until `[OKAY]`; stop instead on
|
|
44
|
+
a material human-owned blocker.
|
|
45
|
+
|
|
46
|
+
These bounded fallbacks apply only to the route, plan-review, and implementation
|
|
47
|
+
questions. Never apply them to secrets, destructive or security-sensitive
|
|
48
|
+
actions, or material human-owned product or architecture decisions.
|
|
29
49
|
|
|
30
50
|
## Ownership
|
|
31
51
|
|
|
@@ -44,7 +64,12 @@ Accelerated the minimum recommendation but does not force it.
|
|
|
44
64
|
migration, concurrency, shared-contract, or high-risk work.
|
|
45
65
|
- Each mutable surface has one writer; independent non-overlapping surfaces may
|
|
46
66
|
split, while coupled surfaces use one `deep` writer and ordered handoffs.
|
|
47
|
-
- Oracle owns `plan-review` when
|
|
67
|
+
- Oracle owns `plan-review` when review is explicitly selected or chosen by the
|
|
68
|
+
bounded recommended fallback. Every route requires
|
|
69
|
+
mandatory verification, but final-verification ownership is proportional:
|
|
70
|
+
trivial deterministic Direct work may be verified by Root when Root is not
|
|
71
|
+
self-approving its own implementation; materially risky Direct work and every
|
|
72
|
+
Accelerated or Full final verify require a fresh read-only Oracle. The
|
|
48
73
|
implementation writer never approves its own work.
|
|
49
74
|
|
|
50
75
|
## Progressive loading
|
|
@@ -90,10 +115,10 @@ node "<skill-dir>/scripts/validate.mjs" --change openspec/changes/<feature> --ro
|
|
|
90
115
|
- **Checklist**: run only when activated; the later `ready` gate includes it if
|
|
91
116
|
present.
|
|
92
117
|
|
|
93
|
-
`ready` validates the artifacts needed before the
|
|
94
|
-
`Review plan with Oracle (Recommended)`
|
|
95
|
-
approval remains separate from implementation confirmation and never
|
|
96
|
-
final verification.
|
|
118
|
+
`ready` validates the artifacts needed before the explicit or bounded-default
|
|
119
|
+
`Review plan with Oracle (Recommended)` / `Proceed without review` choice.
|
|
120
|
+
Review approval remains separate from implementation confirmation and never
|
|
121
|
+
satisfies final verification.
|
|
97
122
|
`closeout` additionally requires completed tasks, independent oracle PASS,
|
|
98
123
|
complete FR/buildable-SC evidence, an observed PASS or explicit residual RISK
|
|
99
124
|
for every outcome SC, and an archive report ready for the transactional archive
|
|
@@ -32,8 +32,11 @@ After deciding delegation creates net gain, select deterministically:
|
|
|
32
32
|
Proven independent surfaces may use different writers only with non-overlapping
|
|
33
33
|
ownership. Overlapping or compatibility-coupled work uses one `deep` writer and
|
|
34
34
|
ordered handoffs. A writer escalates before expanding beyond its routing
|
|
35
|
-
boundary.
|
|
36
|
-
|
|
35
|
+
boundary. Every route requires mandatory verification, and the implementation
|
|
36
|
+
writer never approves its own work. For trivial deterministic Direct work, Root
|
|
37
|
+
may run the proportionate focused checks when Root is not self-approving its own
|
|
38
|
+
implementation. For materially risky Direct work and every Accelerated or Full
|
|
39
|
+
final verify, root hands the unchanged candidate to a fresh read-only Oracle.
|
|
37
40
|
|
|
38
41
|
Implementation evidence may refine an artifact without restarting the workflow:
|
|
39
42
|
|
|
@@ -44,3 +47,11 @@ Implementation evidence may refine an artifact without restarting the workflow:
|
|
|
44
47
|
silently expanding this one.
|
|
45
48
|
|
|
46
49
|
Report files changed, checks executed, deviations, and remaining work.
|
|
50
|
+
|
|
51
|
+
## Native dispatch groups and waves
|
|
52
|
+
|
|
53
|
+
For each ready group, confirm prerequisites, ordered lane tasks, exact path
|
|
54
|
+
union, owner, requirements, and checks. Select the capacity-bounded set of
|
|
55
|
+
undispatched ready lanes and create one fresh native specialist assignment per admitted lane. Issue every dispatch in that native wave before any wait, status,
|
|
56
|
+
result collection, or root implementation action. Retain handles; when terminal evidence frees capacity, dispatch the next undispatched ready lane before waiting again. Validate terminal evidence per lane, while root alone updates task state. Cross the declared barrier only after every lane is terminal and reconciled. Preserve truthful capability/capacity fallback: when native
|
|
57
|
+
concurrency is unavailable or width is one, use a truthful sequential fallback.
|
|
@@ -22,6 +22,27 @@ not justify artificial implementation tasks.
|
|
|
22
22
|
|
|
23
23
|
Order test-first work before its implementation, group tasks by independently
|
|
24
24
|
deliverable story, identify the MVP, and state dependencies. `[P]` is permitted
|
|
25
|
-
only
|
|
26
|
-
|
|
25
|
+
only for tasks assigned to a declared lane whose mutable path union does not
|
|
26
|
+
overlap another lane in the same group. If there is no safe parallel work, use
|
|
27
27
|
`- None: <evidence-backed reason>` instead. Avoid ceremonial tasks.
|
|
28
|
+
|
|
29
|
+
## Parallel execution
|
|
30
|
+
|
|
31
|
+
Declare safe work with this exact grammar:
|
|
32
|
+
|
|
33
|
+
### Group P1
|
|
34
|
+
|
|
35
|
+
- Lane L1: T001 -> T002 | Owner: deep
|
|
36
|
+
- Lane L2: T003 -> T004 | Owner: quick
|
|
37
|
+
- Prerequisites: None
|
|
38
|
+
- Barrier: Final verification
|
|
39
|
+
- Rationale: Both lane path sets are disjoint and neither lane consumes peer output.
|
|
40
|
+
|
|
41
|
+
Groups and lanes are sequential and unique. Every `[P]` task belongs to exactly one lane; non-`[P]` tasks do not. Lane surfaces are the union of exact task
|
|
42
|
+
paths, and lanes in a group have disjoint surfaces and no cross-lane dependency.
|
|
43
|
+
The `Rationale` must explicitly state path-disjointness and cross-lane dependency
|
|
44
|
+
evidence so structural validation can gate the evidence shape and Oracle can
|
|
45
|
+
judge whether the claim is true.
|
|
46
|
+
Prerequisites name known tasks outside the group; the barrier is a downstream
|
|
47
|
+
task or `Final verification` after all lane members. If no group exists, use
|
|
48
|
+
only `- None: <evidence-backed reason>` and no task may use `[P]`.
|
|
@@ -1,10 +1,12 @@
|
|
|
1
1
|
# Verify contract
|
|
2
2
|
|
|
3
|
-
**Owner**:
|
|
3
|
+
**Owner**: route- and risk-aware; Root for trivial deterministic Direct, otherwise a fresh read-only Oracle
|
|
4
4
|
|
|
5
|
-
|
|
6
|
-
|
|
7
|
-
|
|
5
|
+
Every route requires mandatory verification. Trivial deterministic Direct work
|
|
6
|
+
may be verified by Root with focused deterministic checks. Materially risky
|
|
7
|
+
Direct work and Accelerated or Full final verification use a fresh read-only
|
|
8
|
+
Oracle. The implementation writer never approves its own work. Inspect changed
|
|
9
|
+
code and run or validate the smallest sufficient checks.
|
|
8
10
|
|
|
9
11
|
Judge three dimensions separately: **completeness** of accepted scope and
|
|
10
12
|
coverage, **correctness** against behavioral contracts and executed evidence,
|
|
@@ -16,7 +18,7 @@ task.
|
|
|
16
18
|
|
|
17
19
|
Return `pass` or `fail`, compliance matrix, commands/results, stable findings,
|
|
18
20
|
critical issues, warnings, and remediation anchors. For Direct, return the
|
|
19
|
-
verdict in-session. For Accelerated and Full, root persists
|
|
21
|
+
verdict in-session regardless of owner. For Accelerated and Full, root persists Oracle's exact result
|
|
20
22
|
using `<skill-dir>/templates/verify-report.md`, where `<skill-dir>` is the
|
|
21
23
|
directory containing the installed `thoth-sdd/SKILL.md`. Fail routes to converge
|
|
22
24
|
(straight back to implement for Direct); PASS permits closeout and archive for
|