ai-engineering-loop 2.3.0 → 2.4.0

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.
Files changed (39) hide show
  1. package/.agents/skills/ai-engineering-loop/references/loop-risky.md +1 -0
  2. package/CHANGELOG.md +25 -0
  3. package/README.md +1 -0
  4. package/THIRD_PARTY_NOTICES.md +11 -0
  5. package/adapters/spec/LICENSE +202 -0
  6. package/adapters/spec/NOTICE-AEL.md +26 -0
  7. package/adapters/spec/agents/ael-spec-codebase-details.md +71 -0
  8. package/adapters/spec/agents/ael-spec-comprehension.md +71 -0
  9. package/adapters/spec/agents/ael-spec-design.md +65 -0
  10. package/adapters/spec/agents/ael-spec-implementation-readiness.md +64 -0
  11. package/adapters/spec/agents/ael-spec-structural.md +67 -0
  12. package/adapters/spec/agents/ael-spec-style.md +68 -0
  13. package/adapters/spec/agents/ael-spec-synthesizer.md +73 -0
  14. package/adapters/spec/agents/ael-spec-zoom-out.md +76 -0
  15. package/adapters/spec/skills/ael-spec-brainstorm/SKILL.md +265 -0
  16. package/adapters/spec/skills/ael-spec-draft/SKILL.md +315 -0
  17. package/adapters/spec/skills/ael-spec-draft/references/design_doc.md +179 -0
  18. package/adapters/spec/skills/ael-spec-drift/SKILL.md +68 -0
  19. package/adapters/spec/skills/ael-spec-resolve/SKILL.md +188 -0
  20. package/adapters/spec/skills/ael-spec-review/SKILL.md +403 -0
  21. package/adapters/spec/skills/ael-spec-review-loop/SKILL.md +191 -0
  22. package/adapters/spec/standards/ael-spec-execution-instructions.md +36 -0
  23. package/adapters/spec/standards/ael-spec-feature-spec-standard.md +38 -0
  24. package/adapters/spec/standards/ael-spec-heilmeier-standard.md +32 -0
  25. package/adapters/spec/standards/ael-spec-implementation-readiness-standard.md +36 -0
  26. package/adapters/spec/standards/ael-spec-re-implementation-standard.md +39 -0
  27. package/adapters/spec/standards/ael-spec-zoom-out-standard.md +62 -0
  28. package/bin/ai-engineering-loop.js +42 -2
  29. package/core/spec-pack.md +47 -0
  30. package/lib/design-ref.js +178 -0
  31. package/lib/gates.js +1 -0
  32. package/lib/goal-runtime.js +5 -0
  33. package/lib/hosts/doctor.js +16 -2
  34. package/lib/impact-map.js +29 -1
  35. package/lib/receipt.js +12 -1
  36. package/lib/smart-context.js +10 -1
  37. package/lib/sync-hosts.js +25 -2
  38. package/package.json +1 -1
  39. package/schemas/goal-contract.schema.json +87 -11
@@ -0,0 +1,191 @@
1
+ ---
2
+ name: ael-spec-review-loop
3
+ license: Apache-2.0
4
+ disable-model-invocation: true
5
+ description: >-
6
+ Executes the Antagonistic-Review process from the Elephant-Goldfish
7
+ Model (EGM) via an automated multi-agent review loop to audit documents
8
+ against qualitative standards and structural templates, and automatically
9
+ apply approved fixes. Use when reviewing design docs, READMEs, feature
10
+ specifications, implementation plans, or verifying document completeness
11
+ and tone.
12
+ ---
13
+
14
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
15
+
16
+ # Goldfish Review: Loop (Orchestration Driver)
17
+
18
+ Host rules: spawn each Goldfish reviewer as a fresh sibling subagent with your host's subagent tool, as described in `npx ai-engineering-loop docs loop-hosts` (read only your host's section). Personas and standards named below resolve with `npx ai-engineering-loop docs <name>`.
19
+
20
+ ## Overview
21
+
22
+ This skill acts as the central orchestration engine driving the multi-agent
23
+ review loop. It coordinates two distinct agent memory models:
24
+
25
+ * **The Goldfish (Adversarial Auditors)**: Subagents spawned in isolated,
26
+ fresh context windows to catch unstated assumptions without historical bias.
27
+ * **The Elephant (Persistent Resolver)**: The primary agent context holding
28
+ institutional knowledge to evaluate findings and apply approved repairs.
29
+
30
+ The driver manages argument intake, resolves paths for reviewers and standards,
31
+ unconditionally re-parses target line ranges before every iteration, spawns
32
+ concurrent specialized Goldfish auditors based on configuration, and coordinates
33
+ with the Elephant context to resolve findings.
34
+
35
+ ## Delegation
36
+
37
+ A Goldfish runs in a fresh context with no conversation history. Spawn each
38
+ subagent with your harness's subagent tool (per host in `npx ai-engineering-loop docs loop-hosts`; for example Grok CLI `spawn_subagent`, Claude Code `Task` or `Agent`, Codex `spawn_agent`, Antigravity `invoke_subagent`). Each subagent is a persona doc (`npx ai-engineering-loop docs <name>`), and its
39
+ frontmatter `name` (for example `ael-spec-zoom-out`) is the subagent type to request
40
+ where your harness registers named subagents. Otherwise spawn a general-purpose
41
+ subagent and tell it to read that persona file first. Pass the prompt verbatim
42
+ and wait for the report. If your harness cannot nest subagents,
43
+ use direct dispatch instead of the synthesizer. See `npx ai-engineering-loop docs loop-hosts`.
44
+ If your host has no subagent tool, do not stop: run each reviewer one at a time in this session, reading only its persona, the standard, and the document (not the earlier conversation), label every report `Review Method: CONTEXT_ISOLATION_ONLY (not an independent Goldfish review)`, and tell the user.
45
+
46
+ ## Inputs
47
+
48
+ The caller or user provides exactly these parameters:
49
+
50
+ * **`TargetFile`**: Absolute or relative path to target document.
51
+ * **`SemanticTarget`**: Descriptive targeting string (e.g., *"Target the H4
52
+ block with SS-ID 'S1.12'"*), or `full file`.
53
+ * **`Reviewers`**: List of reviewer names (shortnames or paths).
54
+ * **`Standards`**: Mapping of reviewer names to standard names (shortnames or
55
+ paths).
56
+ * **`GoldfishContext`**: A reading list (source files, directories, or
57
+ documentation URLs) fed to reviewers to establish their baseline.
58
+
59
+ ## Path & Standard Resolution Logic
60
+
61
+ When resolving names in `Reviewers` and `Standards`:
62
+
63
+ * **Reviewer Shortnames**: Map reviewer shortname (e.g., `design_critic`,
64
+ `structural`, `style`, `implementation_readiness`,
65
+ `codebase_details`, `comprehension`, `zoom_out`) to `npx ai-engineering-loop docs ael-spec-<persona>`, where `<persona>` is the shortname with `_` written as `-` (`design_critic` is `ael-spec-design`), or an `ael-spec-*` name as is.
66
+ The persona file defines the reviewer's core mandates, audit vectors, and
67
+ output format.
68
+ * **Standard Shortnames**: Map standard shortname (e.g., `feature-spec-standard`,
69
+ `implementation-readiness-standard`, `zoom-out-standard`, `heilmeier-standard`,
70
+ `re-implementation-standard`) to `npx ai-engineering-loop docs ael-spec-{name}`.
71
+ * **Custom Paths**: If the name contains a slash (`/`), treat it as an
72
+ absolute or workspace-relative path and read it directly.
73
+
74
+ ## Execution Workflow
75
+
76
+ ### Step 1: Intake & Configuration UX
77
+
78
+ Evaluate the provided inputs (`TargetFile`, `SemanticTarget`, `Reviewers`,
79
+ `Standards`, `GoldfishContext`).
80
+
81
+ 1. **Identify Missing Inputs**: Determine which of the required parameters are
82
+ missing.
83
+ 2. **Interactive Onboarding**:
84
+ * If **all** inputs are present, proceed to Step 2.
85
+ * If **some or all** inputs are missing, interactively prompt the user
86
+ **only for the missing inputs**, one at a time via chat messages.
87
+ 3. **Reviewer Recommendations**:
88
+ * *Feature Specs*: Recommend `design_critic` (mapped to
89
+ `npx ai-engineering-loop docs ael-spec-feature-spec-standard`) and
90
+ `structural` (mapped to `npx ai-engineering-loop docs ael-spec-feature-spec-standard`).
91
+ * *Implementation Plans*: Recommend `implementation_readiness`
92
+ (mapped to `npx ai-engineering-loop docs ael-spec-implementation-readiness-standard`)
93
+ and `codebase_details`.
94
+ 4. **Confirm Configuration**: Display the consolidated plan (Target, Reviewers,
95
+ and mapped Standards) and ask the user to confirm before proceeding.
96
+
97
+ ### Step 2: The Sharpening Loop (Iterative Cycle)
98
+
99
+ Execute the following sequence in a continuous loop:
100
+
101
+ #### 2.1 Unconditional Target Resolution
102
+
103
+ Before every iteration, unconditionally re-scan `TargetFile` by reading it
104
+ and using the stable `SemanticTarget` string (e.g., locating the target header) to
105
+ dynamically determine the updated absolute start and end line numbers. This
106
+ ensures absolute accuracy even if the file was edited in prior rounds. (If `full
107
+ file` mode is specified, confirm full-file line bounds).
108
+
109
+ #### 2.2 Review Swarm Execution
110
+
111
+ 1. **Launch Subagent Swarm**:
112
+ Spawn subagents to execute the review:
113
+ * **Hierarchical Dispatch (Synthesizer)**: Spawn the intermediate
114
+ synthesizer subagent (`npx ai-engineering-loop docs ael-spec-synthesizer` / `ael-spec-synthesizer`).
115
+ The synthesizer spawns the Goldfish review swarm in
116
+ parallel, deduplicates findings, writes the report artifact, and
117
+ returns a single summary message to avoid context pollution in the root
118
+ conversation. Pass it the Failed Reviewers rule below. Its summary must
119
+ give a status, `completed` or `not run`, for each requested reviewer.
120
+ * **Direct Dispatch (Flat Swarm)**: Alternatively, spawn each requested
121
+ reviewer subagent in `Reviewers` directly.
122
+ Instruct each subagent to first read its persona definition in
123
+ `npx ai-engineering-loop docs ael-spec-<persona>`, where `<persona>` is the shortname with `_` written as `-` (`design_critic` is `ael-spec-design`), or an `ael-spec-*` name as is, then audit `TargetFile` at the resolved line range
124
+ against its designated standard in the `ael-spec-*-standard` docs,
125
+ supplying `GoldfishContext`.
126
+ * Each reviewer produces findings with severity (🔴 Blocker, 🟡 Major, 🟢
127
+ Nit), line numbers, discrepancy descriptions, and rationales.
128
+ 2. **Consolidate Findings**: Collect all findings across reviewers into a single
129
+ flat `FindingsList`. Give each finding a stable id
130
+ (`<reviewer>-<round>-<n>`) and record its reviewer, severity, and affected
131
+ scope.
132
+ 3. **Failed Reviewers**: A reviewer that times out, fails, or returns no report
133
+ did not review. Retry it once. If it fails again, record it as **not run**
134
+ in the report. Never count it as a reviewer with zero findings. If the
135
+ synthesizer summary has no status for a requested reviewer, record that
136
+ reviewer as not run.
137
+
138
+ #### 2.3 Aggregated Resolution Hand-off
139
+
140
+ 1. Pass `TargetFile`, the active line range, and the aggregated `FindingsList`
141
+ to the **`ael-spec-resolve`** workflow.
142
+ 2. Execute the resolution workflow in the primary Elephant context:
143
+ * Categorize findings into `CONTEST` vs. `PROPOSE FIX`.
144
+ * Save `goldfish_review_report.md` with your file-writing tool.
145
+ * Present the interactive decision gate (Auto accept, Continue, Interactive
146
+ breakdown).
147
+ * Apply approved text repairs to `TargetFile` using your file-editing tool.
148
+
149
+ #### 2.4 Iteration Check & Turn Boundary
150
+
151
+ Fresh reviewers sample differently each round, so a round with zero findings
152
+ is rare even when the document improves. Use these rules instead:
153
+
154
+ * **Settled findings**: A finding is settled when it is contested or
155
+ recorded without a fix (accepted risk). Keep the settled list across
156
+ rounds. If a new reviewer raises a settled finding again (same affected
157
+ scope, same issue), do not re-open it.
158
+ * **Fixed findings**: An applied fix does not settle a finding. If a later
159
+ reviewer raises the same issue at the same scope, the fix failed or a
160
+ later edit reintroduced the issue. Re-open the finding.
161
+ * **Clean round**: A round is clean when no reviewer raises a Blocker or
162
+ Major finding that is not settled. Nits do not block. A reviewer recorded
163
+ as not run makes the round not clean. Decide this from the review results,
164
+ before you apply the fixes of that round.
165
+ * **Converged**: Two consecutive clean rounds. One clean round is not enough,
166
+ because the next fresh reviewer can find a different gap in the same text.
167
+ * **Cap**: 5 rounds.
168
+ * **Churn**: A new finding is at the scope of an applied fix, and reverting
169
+ that fix would resolve it.
170
+
171
+ On converged, go to Step 3. On cap or churn, write a short list of the open
172
+ findings by severity, then ask the user one question: accept and proceed,
173
+ continue past the cap, or return the document to drafting.
174
+
175
+ **Crucial Turn-Taking Requirement**: After the resolution phase finishes for a
176
+ given round, the loop driver **MUST respect the turn-taking checkpoint in
177
+ `ael-spec-resolve`**. The driver must remain idle across human turns,
178
+ only looping back to Step 2.1 when the user explicitly confirms they wish to run
179
+ another round of review.
180
+
181
+ ### Step 3: Output Boundary & Final Checkpoint
182
+
183
+ Upon completion (converged, cap or churn accepted by the user, or manual
184
+ halt):
185
+
186
+ * **Programmatic Callers**: If running as a subflow of `ael-spec-review`, exit
187
+ cleanly to hand execution back to the parent workflow.
188
+ * **Manual Invocations**: Output a summary of completion in chat: *"The review
189
+ loop has completed after [N] rounds ([converged | accepted at cap | halted]).
190
+ Open findings are in the report. Would you like to proceed to
191
+ the next stage, or do you have any final questions?"*
@@ -0,0 +1,36 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ ## Execution Instructions
4
+
5
+ * **Design-Only Scope**: You are strictly forbidden from
6
+ writing, modifying, or building any production implementation files (e.g.,
7
+ `.h`, `.cc`, `.py`, `.go`) in the source tree. You may only write code
8
+ inside a markdown draft block or in a temporary scratchpad file.
9
+ * **Scratchpad Location (CRITICAL)**: You MUST write any temporary
10
+ scratchpad code files ONLY outside the project source tree (e.g.,
11
+ under `/tmp/` or your conversation's active scratch directory). You
12
+ are STRICTLY FORBIDDEN from creating `.h`, `.cc`, `.py`, `.go`, or any
13
+ other code files anywhere inside the project source tree.
14
+ * **HOW TO RESPOND**: If the user asks for production implementation while
15
+ this design workflow is active, explain that it produces design documents
16
+ and illustrative code only. Label any markdown or scratchpad example
17
+ clearly. Do not imply that production files were changed. Ask whether to
18
+ continue the design or leave the workflow for implementation.
19
+ * **Explicit Consensus Gates**: You must confirm with the user and achieve
20
+ explicit agreement at each step transition. Do not save the document or move
21
+ to the next step without the user's explicit "go-ahead". Do not move on
22
+ easily. Challenge the user at every step.
23
+ * **ONE QUESTION AT A TIME (CRITICAL)**: You are strictly forbidden from
24
+ asking more than one question in a single response. You must present exactly
25
+ one question or request for clarification, wait for the user's response, and
26
+ only then proceed. Do not bundle questions or present a list of questions
27
+ for the user to answer at once.
28
+ * **NO SUMMARIES (CRITICAL)**: You MUST NOT include a "Summary of work",
29
+ "Status", "Adherence", or "Rule checklist" in your final response. Respond
30
+ only with the technical content and next steps.
31
+ * **Contradiction Auditing (MUST ENFORCE - CRITICAL)**: If you detect a
32
+ contradiction between the user's new requests and earlier
33
+ goals/architecture, **you MUST NOT modify any files**. You must only respond
34
+ to the user in plain text, point out the contradiction, and ask for
35
+ resolution. You are strictly forbidden from planning or executing any
36
+ file-writing/editing tool in the same turn as you discover a contradiction.
@@ -0,0 +1,38 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ # Feature Specification Standard
4
+
5
+ This standard defines the quality bar for high-level feature specifications
6
+ before implementation begins.
7
+
8
+ ## Core Principle
9
+
10
+ A feature specification passes if a competent engineer (or fresh AI session) who
11
+ is unfamiliar with the codebase can understand the proposed feature, the
12
+ business motivation, and the design trade-offs, and confirm that all major risks
13
+ and alternatives have been addressed.
14
+
15
+ ## The Alternatives Rule
16
+
17
+ The specification must include a **mandatory evaluation of alternatives**. It
18
+ fails this standard if it proposes a single path without proving that other
19
+ viable designs were considered and rejected for specific, documented reasons
20
+ (including, but not limited to, complexity, cost, performance, or scalability).
21
+
22
+ ## Risks & Limits
23
+
24
+ The specification must explicitly call out its own constraints. It must identify
25
+ potential failure modes, security implications, scalability limits, and how they
26
+ will be mitigated.
27
+
28
+ ## Requirements (Non-Exhaustive Examples)
29
+
30
+ The specification must detail these areas if applicable to the changes:
31
+
32
+ * **Problem Statement**: The business or user problem being solved.
33
+ * **Goals & Non-Goals**: What the feature will and will not deliver.
34
+ * **Proposed Architecture**: Component layout, data flows, and ecosystem
35
+ integration.
36
+ * **Alternatives Considered**: Rejected options and rationale for rejection.
37
+ * **Risks & Mitigations**: Identified security, scalability, or complexity
38
+ risks.
@@ -0,0 +1,32 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ # Heilmeier Catechism Standard
4
+
5
+ This standard defines the quality bar for project proposals based on the
6
+ Heilmeier Catechism.
7
+
8
+ ## Core Principle
9
+
10
+ A proposal passes if a competent engineer (or fresh AI session) who is
11
+ unfamiliar with the project can understand the goals, risks, and value of the
12
+ proposed work, and confirm that all catechism questions are answered without
13
+ ambiguity.
14
+
15
+ ## Requirements (The Catechism Questions)
16
+
17
+ To satisfy this standard, the proposal must explicitly and clearly answer the
18
+ following questions:
19
+
20
+ * **Objectives**: What is being attempted? (Must be written with absolutely no
21
+ jargon).
22
+ * **State of the Art**: How is it done today, and what are the limits of
23
+ current practice?
24
+ * **New Approach**: What is new in the approach, and why is it expected to be
25
+ successful?
26
+ * **Impact**: Who cares? If successful, what difference will it make?
27
+ * **Risks**: What are the risks and the payoffs?
28
+ * **Cost**: How much will it cost in terms of resources, compute, or
29
+ headcount?
30
+ * **Timeline**: How long will it take?
31
+ * **Metrics & Exams**: What are the mid-term and final "exams" to check for
32
+ success?
@@ -0,0 +1,36 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ # Implementation Readiness Standard
4
+
5
+ This standard defines the quality bar for implementation plans and task lists.
6
+
7
+ ## Core Principle
8
+
9
+ A plan passes if a competent engineer (or fresh AI session) who is unfamiliar
10
+ with the codebase can execute the plan and produce the correct changes on the
11
+ first attempt, without guessing, asking clarifying questions, or making
12
+ independent design choices.
13
+
14
+ ## The Zero-Decision Rule
15
+
16
+ The plan must leave **zero design or architectural choices to the implementer**.
17
+ If the implementer must make design choices (including, but not limited to,
18
+ choosing a database schema, designing an API path, selecting a dependency, or
19
+ deciding how to handle a new edge case), the plan **fails** this standard. All
20
+ such decisions must be pre-determined and documented in the plan.
21
+
22
+ ## Phasing & Checkpoints
23
+
24
+ The plan must partition the work into logical, incremental phases. Each phase
25
+ must leave the codebase in a stable, buildable, and testable state.
26
+
27
+ ## Requirements (Non-Exhaustive Examples)
28
+
29
+ The plan must detail these areas if applicable to the proposed changes:
30
+
31
+ * **Target Scope**: Explicit list of files to be created, modified, or
32
+ deleted.
33
+ * **Actionable Steps**: Precise logic changes, data transformations, and API
34
+ updates for each file.
35
+ * **Test Plan**: Specific test cases (happy path and edge cases) that must be
36
+ implemented.
@@ -0,0 +1,39 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ # Re-Implementation Standard
4
+
5
+ This standard defines the quality bar for documenting codebase state.
6
+
7
+ ## Core Principle
8
+
9
+ A document passes if a competent engineer (or fresh AI session) who is
10
+ unfamiliar with the codebase can write functionally identical code on the first
11
+ attempt using *only* the document, without guessing or making independent design
12
+ choices.
13
+
14
+ ## The Zero-Assumption Rule
15
+
16
+ The document fails if the reader must guess parameter defaults, assume error
17
+ handling, choose design patterns, or infer output schemas. Every behaviorally
18
+ significant contract must be explicit.
19
+
20
+ ## Scope of Detail
21
+
22
+ * **Include**: External contracts (APIs, formats, side-effects), behavioral
23
+ rules (logic, state, errors), and performance constraints.
24
+ * **Exclude**: Syntax (loops, variable names), internal code organization
25
+ (private helpers), and non-observable structures (internal collections).
26
+
27
+ ## Requirements (Non-Exhaustive Examples)
28
+
29
+ Document these areas if applicable to the codebase scope:
30
+
31
+ * **APIs**: Signatures, types, and expected behavior.
32
+ * **Logic**: Control flow, algorithms, and validation.
33
+ * **Data**: Schemas, serialization, and state transitions.
34
+ * **Side Effects**: Database writes, network calls, logging, and metrics.
35
+ * **Dependencies**: System integrations and libraries.
36
+ * **Configuration**: Flags, parameters, and hooks.
37
+ * **Constraints**: Hardcoded limits and assumptions.
38
+ * **Errors**: Edge cases, empty states, and retries.
39
+ * **Tests**: Test helpers and mocks (if documenting tests).
@@ -0,0 +1,62 @@
1
+ <!-- Generated by scripts/sync-spec-pack.js from egagofur/elephant-goldfish-model@228457d1e0b6. Derived from the Elephant-Goldfish Model, Copyright Google LLC, Apache-2.0; modified: renamed to ael-spec-*, host-neutral references. Do not edit; run the script. -->
2
+
3
+ # Zoom Out Standard
4
+
5
+ This standard defines the quality bar for evaluating high-level designs,
6
+ architectures, and implementation plans from a high-altitude perspective.
7
+
8
+ ## Core Principle
9
+
10
+ A design passes if it solves the root problem in the simplest possible way,
11
+ leverages existing infrastructure and codebase conventions, avoids premature
12
+ optimization or speculative features (YAGNI), and maintains appropriate
13
+ abstraction boundaries without getting bogged down in low-level code
14
+ implementation details.
15
+
16
+ ## 0. Context Sufficiency (Prerequisite Check)
17
+
18
+ * **Understanding the Objective**: The document must contain enough
19
+ information to define the overall objective. If the reviewer cannot
20
+ determine *what* we are trying to achieve and *why*, the review must fail
21
+ immediately with a request for context.
22
+
23
+ ## 1. Root Problem Focus
24
+
25
+ * **Symptom vs. Root Cause**: The design must solve the underlying problem,
26
+ not just apply a local band-aid to symptoms.
27
+ * **Project Alignment**: The design must align with the high-level goals and
28
+ architectural philosophy defined in the project's root documentation
29
+ (`README.md` or `context.md`).
30
+
31
+ ## 2. Simplicity & Extensibility
32
+
33
+ * **YAGNI (You Aren't Gonna Need It)**: Reject abstractions, code paths, or
34
+ features built solely for speculative future requirements.
35
+ * **Premature Optimization**: Avoid introducing complexity (caching, sharding,
36
+ async queues) before performance bottlenecks are proven.
37
+ * **Minimalist Data Flow**: Ensure data passes through the minimal number of
38
+ components.
39
+ * **No Architectural Dead-Ends**: Use clean interfaces and boundaries to allow
40
+ future expansion without blocking it, without implementing speculative
41
+ features now.
42
+
43
+ ## 3. Anti-Weeds (Altitude Control)
44
+
45
+ * **Premature Detail**: High-level sections must focus on architecture,
46
+ interfaces, and data flows, not code-level implementation details (e.g.,
47
+ helper functions or loop variables).
48
+ * **Workarounds vs. Structural Fixes**: Avoid complex local workarounds;
49
+ propose changes to the root component if that is cleaner.
50
+ * **Right Placement (Architectural Fit)**: Ensure new logic is placed in the
51
+ correct layer/component according to separation of concerns.
52
+
53
+ ## 4. Leverage Existing
54
+
55
+ * **Codebase Reuse**: Search the codebase and reuse existing utilities,
56
+ classes, or modules instead of reinventing them. Ensure the candidate for
57
+ reuse is stable, maintained, and does not violate dependency boundaries.
58
+ * **Pattern Consistency**: Follow established design patterns in the codebase.
59
+ * **Platform Standards**: Use established ecosystem frameworks and standard libraries
60
+ instead of bespoke or ad-hoc custom solutions.
61
+ * **Identify Anti-Patterns**: Actively flag known anti-patterns
62
+ (e.g., bypassing standard authentication, hardcoding deployment configurations).
@@ -1477,6 +1477,7 @@ function handleRun() {
1477
1477
  return { stale: h.stale, staleFiles: h.staleFiles, installedVersion: h.installedVersion, packageVersion: VERSION };
1478
1478
  })(),
1479
1479
  learnings: runLearnings(),
1480
+ ...specHintField(runResult),
1480
1481
  studioUrl: readStudioUrl(CWD)
1481
1482
  }, null, 2));
1482
1483
  return;
@@ -1486,6 +1487,8 @@ function handleRun() {
1486
1487
  const t = runResult.run.tier;
1487
1488
  console.log(`- Tier: ${t.current} (${t.source}; ${t.reasons.map((r) => r.code).join(', ')})`);
1488
1489
  }
1490
+ const hint = specHintField(runResult).specHint;
1491
+ if (hint) console.log(`- Spec: ${hint}`);
1489
1492
  console.log(`- State: ${runResult.run.state}, iteration ${runResult.run.iteration}`);
1490
1493
  console.log(`- Mode: ${runResult.run.mode || RUN_MODES.ASSISTED}`);
1491
1494
  if (runResult.run.workflow) {
@@ -3055,7 +3058,7 @@ function requiredIntegerArg(name) {
3055
3058
 
3056
3059
  function readBoundedJsonInput(fileName) {
3057
3060
  if (!fileName) {
3058
- const error = new Error('goal draft requires --file <json>');
3061
+ const error = new Error('goal draft requires --file <json> or --from-design <doc.md>');
3059
3062
  error.code = 'INVALID_USAGE';
3060
3063
  throw error;
3061
3064
  }
@@ -3075,6 +3078,23 @@ function readBoundedJsonInput(fileName) {
3075
3078
  return JSON.parse(fs.readFileSync(file, 'utf8'));
3076
3079
  }
3077
3080
 
3081
+ // Spec pack hint (#99): once, when a new Run starts at T2. T0/T1 and resumed
3082
+ // Runs get nothing, so their output is unchanged.
3083
+ function specHintField(runResult) {
3084
+ const run = runResult.run;
3085
+ if (runResult.resumed || run.tier?.current !== 'T2') return {};
3086
+ let installed = false;
3087
+ try {
3088
+ installed = hostSkillStatus({ home: homeDir(), version: VERSION }).packs.includes('spec');
3089
+ } catch {
3090
+ installed = false;
3091
+ }
3092
+ return {
3093
+ specHint: 'For a new feature or architecture change, design first: ael-spec-brainstorm -> ael-spec-draft -> ael-spec-review, then `goal draft --from-design docs/designs/<name>.md`.' +
3094
+ (installed ? '' : ' Install with `sync-hosts --pack spec`.')
3095
+ };
3096
+ }
3097
+
3078
3098
  function handleGoal() {
3079
3099
  const action = args[1] || 'show';
3080
3100
  try {
@@ -3083,6 +3103,7 @@ function handleGoal() {
3083
3103
  const extra = {};
3084
3104
  if (action === 'show') {
3085
3105
  result = loadGoalDraft(CWD, runId);
3106
+ if (result?.contract?.designRef) extra.design = require('../lib/design-ref.js').designStatus(CWD, result.contract.designRef);
3086
3107
  } else if (action === 'draft') {
3087
3108
  appendRunLifecycle(CWD, runId, {
3088
3109
  type: 'GOAL_DRAFTING',
@@ -3090,7 +3111,24 @@ function handleGoal() {
3090
3111
  actor: argValue('--by') || 'agent',
3091
3112
  message: argValue('--message') || 'Updating the Goal draft'
3092
3113
  });
3093
- const draftGoal = readBoundedJsonInput(argValue('--file'));
3114
+ const designFile = argValue('--from-design');
3115
+ let draftGoal;
3116
+ if (designFile) {
3117
+ const designRef = require('../lib/design-ref.js');
3118
+ const doc = designRef.readDesignDoc(CWD, designFile);
3119
+ const parsed = designRef.parseDesignDoc(doc.text);
3120
+ draftGoal = argValue('--file')
3121
+ ? { ...readBoundedJsonInput(argValue('--file')), designRef: designRef.designRefFrom(doc, parsed) }
3122
+ : designRef.draftFromDesign(doc, parsed);
3123
+ extra.design = {
3124
+ path: doc.path,
3125
+ targetFiles: parsed.targetFiles.length,
3126
+ placeholders: designRef.placeholderRows(draftGoal),
3127
+ warnings: parsed.warnings
3128
+ };
3129
+ } else {
3130
+ draftGoal = readBoundedJsonInput(argValue('--file'));
3131
+ }
3094
3132
  result = saveGoalDraft(CWD, runId, draftGoal, {
3095
3133
  actor: argValue('--by') || 'agent',
3096
3134
  expectedRevision: requiredIntegerArg('--revision')
@@ -3626,6 +3664,7 @@ function handleReviewPack() {
3626
3664
  if (result.impactDrift?.outOfMap.length) {
3627
3665
  console.log(`- Outside the impact map: ${result.impactDrift.outOfMap.join(', ')} (check their side effects)`);
3628
3666
  }
3667
+ for (const line of result.impactDrift?.design || []) console.log(`- ${line}`);
3629
3668
  }
3630
3669
  } catch (err) {
3631
3670
  stageJsonError(err, json, 'REVIEW_PACK_FAILED');
@@ -3969,6 +4008,7 @@ Commands:
3969
4008
  goal Share one revision-bound Goal draft between the Agent and Studio
3970
4009
  show [--run <id>]
3971
4010
  draft --file <json> --revision <n> [--run <id>] [--by <agent>]
4011
+ draft --from-design <doc.md> [--file <json>] --revision <n> prefill from a design doc, record designRef
3972
4012
  freeze --revision <n> --hash <sha256> [--run <id>] [--by <name>]
3973
4013
  freeze --latest [--run <id>] --by <name> freeze the newest draft (CLI resolves revision/hash)
3974
4014
  verification Summarize verification.json for compact Judge context
@@ -0,0 +1,47 @@
1
+ # Spec pack: design before code
2
+
3
+ The optional `spec` host pack adds design-document skills derived from Google's
4
+ [Elephant-Goldfish Model](https://github.com/google/elephant-goldfish-model) (EGM,
5
+ Apache-2.0) through the fork [egagofur/elephant-goldfish-model](https://github.com/egagofur/elephant-goldfish-model).
6
+ The Elephant (your session) drafts; fresh Goldfish reviewers read only the document and
7
+ the code it cites; you decide the trade-offs.
8
+
9
+ ## When to use it
10
+
11
+ - Use it for a new feature, an architecture change, an API or integration design, or a
12
+ data-model change: work that needs decisions before code. A new T2 Run suggests it.
13
+ - Do not use it for a bug fix, typo, small refactor, or any T0/T1 task. Use the normal loop.
14
+
15
+ ## Flow
16
+
17
+ 1. `ael-spec-brainstorm <idea>`: clarifying questions, then six review lenses; writes `concepts_brief.md`.
18
+ 2. `ael-spec-draft`: section-by-section `design_draft.md`, saved under `docs/designs/<name>.md`.
19
+ 3. `ael-spec-review docs/designs/<name>.md`: staged Goldfish review (zoom out, comprehension,
20
+ critique, readiness). Expect many findings; take what is relevant.
21
+ 4. `ael-spec-resolve`: each finding is PROPOSE FIX, CONTEST, ESCALATE or DECLINE; you choose.
22
+ 5. You approve and commit the design. Agreement in chat is not approval.
23
+ 6. `npx ai-engineering-loop goal draft --run <id> --from-design docs/designs/<name>.md --revision <r>`:
24
+ prefilled acceptance rows with `<fill: …>` placeholders, `outOfScope`, and `designRef`.
25
+ Fill the placeholders; `goal freeze` refuses them. Then run the loop as usual.
26
+ 7. Verify and review report design drift (planned files untouched, changed files outside
27
+ the plan). Later, `ael-spec-drift docs/designs/` compares all designs with the code.
28
+
29
+ Personas and standards: `npx ai-engineering-loop docs ael-spec-zoom-out`,
30
+ `docs ael-spec-implementation-readiness-standard`, and so on.
31
+
32
+ ## Hosts
33
+
34
+ `npx ai-engineering-loop sync-hosts --pack spec` installs the six `ael-spec-*` skills on
35
+ every enabled host. Reviewers spawn as fresh sibling subagents per `docs loop-hosts`. A host
36
+ without a subagent tool runs the reviewers one at a time in-session and labels the reports
37
+ `CONTEXT_ISOLATION_ONLY`; that is weaker than a Goldfish review and is disclosed.
38
+ `hosts doctor` shows, per host, whether the pack is installed and which mode it gets.
39
+
40
+ ## Updating from upstream
41
+
42
+ 1. `gh repo sync egagofur/elephant-goldfish-model` (pulls google/main into the fork).
43
+ 2. In this repository: `node scripts/sync-spec-pack.js --pull [ref]`, then `npm test`.
44
+ 3. Open a PR. Users get it through the normal CLI update.
45
+
46
+ The pinned commit is in `vendor/egm/UPSTREAM.json`; every generated file names it.
47
+ Attribution and the list of modifications: `THIRD_PARTY_NOTICES.md` and `adapters/spec/NOTICE-AEL.md`.