@difflab/difflab-cli 0.11.0 → 0.13.0

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 (3) hide show
  1. package/README.md +8 -4
  2. package/dist/index.js +157 -109
  3. package/package.json +1 -1
package/README.md CHANGED
@@ -66,7 +66,7 @@ difflab-cli templates list
66
66
  difflab-cli templates scaffold spec-driven-plan docs/plans first-plan.md --cwd /absolute/project/path
67
67
  ```
68
68
 
69
- The CLI defaults `--cwd` to its current directory. The MCP server exposes `template_list({})` and `scaffold({"name":"spec-driven-plan","cwd":"/absolute/project/path","path":"docs/plans","filename":"first-plan.md"})`. Both interfaces take a hyphenated template name, a relative destination **directory** (`path`, or `.` for `cwd`), and a separate **filename**. MCP always requires an absolute existing `cwd`. Missing destination directories are created. Available names: `spec-driven-plan`, `software-architecture-design`, `architecture-decision-record`, `product-requirements-document`, `agents`, `code-review`, `code-review-rubric`, `planning-intent`, `ui-component-architecture`, `pull-request-description` (adapted from Diffpi's PR template), `flow-definition`, `flow-instance`, `exploration-summary`, `exploration-research`, `exploration-proposal`, and `poc-readme`.
69
+ The CLI defaults `--cwd` to its current directory. The MCP server exposes `template_list({})` and `scaffold({"name":"spec-driven-plan","cwd":"/absolute/project/path","path":"docs/plans","filename":"first-plan.md"})`. Both interfaces take a hyphenated template name, a relative destination **directory** (`path`, or `.` for `cwd`), and a separate **filename**. MCP always requires an absolute existing `cwd`. Missing destination directories are created. Available names: `spec-driven-plan`, `software-architecture-design`, `architecture-decision-record`, `product-requirements-document`, `agents`, `code-review`, `code-review-rubric`, `planning-intent`, `ui-component-architecture`, `pull-request-description` (adapted from Diffpi's PR template), `flow-definition`, `flow-instance`, `exploration-summary`, `exploration-research`, `exploration-proposal`, `poc-readme`, `hook-definition`, and `task`.
70
70
 
71
71
  Reviews scaffold `code-review` as `REVIEW.md` and `code-review-rubric` as `rubric.yaml` in `.difflab/reviews/YYMMDD-name`. The installed rubric is `~/.difflab/templates/code-review-rubric.yaml`. Fill its criterion results during the review. File-scoped criteria require a path-tagged result for every changed source file. Function findings can include `file:line` locations. Evidence guidance and review instructions stay in the review skill, outside the lean YAML template.
72
72
 
@@ -74,11 +74,15 @@ Use `difflab-cli templates scaffold agents . AGENTS.md --cwd /absolute/project/p
74
74
 
75
75
  On first use, missing bundled templates and filled examples are copied to `~/.difflab/templates` and `~/.difflab/templates/examples`. Edit the installed templates to customize future documents. Later runs add missing files but never replace existing ones, even after an upgrade. Examples are for reference, not selectable templates. Copying is literal: prompts and placeholders are not substituted. An unknown name, missing or invalid source, unsafe path, symlink, or existing destination causes an error; destination files are never overwritten. The target directory must remain inside `cwd` and the filename must be a single file name without separators. The one symlink exception is an initialized repository's `.difflab` link to its own user project store: `templates scaffold spec-driven-plan .difflab/plans/YYMMDD-name PLAN.md --cwd /absolute/repo` works in Git worktrees as well. To restore a bundled default, move your installed copy aside before listing again.
76
76
 
77
- The portable `skills/difflab-plan`, `skills/difflab-review`, `skills/difflab-flow`, `skills/difflab-explore`, `skills/difflab-adr`, and `skills/difflab-poc` directories contain natural-language agent workflows for these templates. Install each skill in your client's Agent Skills directory. The MCP-only `log_append({"path":"/absolute/physical/project-store/plans/YYMMDD-name/logs.txt","message":"Phase started"})` tool creates an append-only progress log on first use; pass the physical path, not the `.difflab` symlink. There is no log CLI command.
77
+ The portable `skills/difflab-plan`, `skills/difflab-review`, `skills/difflab-flow`, `skills/difflab-hooks`, `skills/difflab-bdd`, `skills/difflab-writing-style`, `skills/difflab-explore`, `skills/difflab-adr`, and `skills/difflab-poc` directories contain natural-language agent workflows for these templates. Install each skill in your client's Agent Skills directory. The MCP-only `log_append({"path":"/absolute/physical/project-store/plans/YYMMDD-name/logs.txt","message":"Phase started"})` tool creates an append-only progress log on first use; pass the physical path, not the `.difflab` symlink. There is no log CLI command.
78
78
 
79
- The [plan authoring guide](../../skills/difflab-plan/references/authoring.md) requires component/process/artifact visuals, compact interfaces, and a nested repo tree at the start of `Consequences`. Add a `Testing` outline of exact end-to-end and integration scenarios. Keep expected commit and green-CI gates, including for `plan go --commit`. That flag permits scoped commits. Push requires separate authorization. Write `N/A.` when infrastructure stays unchanged. Source examples live in the skill's `references/examples/` and are not CLI templates.
79
+ The [plan authoring guide](../../skills/difflab-plan/references/authoring.md) applies `difflab-writing-style` and `difflab-bdd`. It requires component/process/artifact visuals, compact interfaces, and a directory-grouped `[M/A/D/R]` repo tree at the start of `Consequences`. Add a `Testing` subsection with a behavior flowchart and a matching `[M/A/D/R]` BDD test tree (`given`/`when`/`then`) mapped to existing suites or clearly labeled manual acceptance checks. Keep expected commit and green-CI gates, including for `plan go --commit`. That flag permits scoped commits. Push requires separate authorization. Write `N/A.` when infrastructure stays unchanged. Source examples live in the skill's `references/examples/` and are not CLI templates.
80
80
 
81
- The [review rubric](../../skills/difflab-review/references/rubric.md) follows language conventions, codebase patterns, and applicable `AGENTS.md` rules. Refactor unclear code at the scope needed. Review application behavior first, then module public APIs, then focused unit behavior. Require clear Arrange/Act/Assert blocks and useful fixtures. Untested private functions alone are not findings. Keep comments short and use simple English. Omit nitpicks and minor optimizations. Explicit security/performance focuses deepen those audits.
81
+ The [review rubric](../../skills/difflab-review/references/rubric.md) follows written `AGENTS.md` rules, `difflab-writing-style`, and `difflab-bdd`. Existing patterns are context, not requirements. Flag material antipatterns even when nearby code shares them. Refactor unclear code at the scope needed. Prefer outside-in tests (`e2e` > `integration` > `unit`) with human-verifiable outputs, and scrutinize test suites against the component's behavior flowchart to propose BDD tree reorganizations (`given`/`when`/`then` with hoisted fixtures) when needed. Untested private functions alone are not findings. Keep whole reports and comments short (one sentence per comment ideally) and use simple English; prefer code snippets or diagrams over walls of text. State findings once and keep detailed coverage in YAML. Omit nitpicks and minor optimizations. Explicit security/performance focuses deepen those audits.
82
+
83
+ Reviews run in the current single agent without subagents or councils, keeping the selected frontier model regardless of vendor. `--adversarial` challenges all dimensions, including Security.
84
+
85
+ PR creation through `pr new`, create, open, or draft always scaffolds and fills the installed `pull-request-description` template before creating the request, then reads the saved body back. `pr merge` watches post-merge CI and then invokes `~/.difflab/hooks/on-merge/HOOK.md` via `difflab-hooks` when registered. `pr merge --squash` selects a native forge squash merge before the operation. These are Agent Skill requests, not CLI subcommands.
82
86
 
83
87
  For an upgrade, reinstall changed skills and reload the client. Back up and move aside only the installed templates/examples you want refreshed, then list templates again. For these changes, inspect `~/.difflab/templates/{spec-driven-plan,code-review}.md` and `examples/{plan-feature,plan-migration,review-change}.md`. Preserve customizations in backups and merge deliberately; never reset the whole directory. Older templates still work: skills add missing subsections to generated documents while preserving human content and existing phase completion state.
84
88
 
package/dist/index.js CHANGED
@@ -48362,7 +48362,7 @@ var spec_driven_plan_default = `# [Outcome to deliver]
48362
48362
 
48363
48363
  ### Big Ideas
48364
48364
 
48365
- [Show changed component relationships with a component diagram, new processes with flowcharts, and new file-based artifacts with trees or suitable visuals. Use Mermaid and fenced text where useful. Keep decisions, responsibilities, invariants, and important failure behavior beside the relevant visual. Draw only the affected parts; keep enough detail to execute without the original chat.]
48365
+ [Lead with compact diagrams. Minimize prose and do not repeat the visual. Use short sentences and familiar words. Show changed component relationships with a component diagram, new processes with flowcharts, and new file-based artifacts with trees or suitable visuals. Use Mermaid and fenced text where useful. Keep decisions, responsibilities, invariants, and important failure behavior beside the relevant visual. Draw only the affected parts; keep enough detail to execute without the original chat.]
48366
48366
 
48367
48367
  ### Infrastructure Changes
48368
48368
 
@@ -48376,11 +48376,11 @@ var spec_driven_plan_default = `# [Outcome to deliver]
48376
48376
 
48377
48377
  #### Repo Changes
48378
48378
 
48379
- [Start with a selected file tree: + new, ~ changed, - removed, * largest edits. Include new modules and meaningful structural changes. Use real inspected paths; label proposed files. State when nothing is removed.]
48379
+ [Start with a directory-grouped file tree using \`[M]\` modified, \`[A]\` added, \`[D]\` deleted, and \`[R]\` renamed before changed entries, including tests. Apply \`/difflab-writing-style\`. Identify the largest edits in a short note. Use real inspected paths and label proposed files. State when nothing is removed.]
48380
48380
 
48381
48381
  #### Testing
48382
48382
 
48383
- [Show a nested outline of end-to-end and integration test scenarios, grouped by feature and behavior. Name the entry point, input/action, and expected result for each scenario. Reuse existing fixtures and suites. Match the scenarios to requirements and phase checks.]
48383
+ [Apply \`/difflab-bdd\`. First draw a Mermaid behavior flowchart showing context branches, state transitions, and converging intermediate states from the user's or caller's perspective. Then map the flowchart into a nested \`[M/A/D/R]\` BDD test tree starting with \`given <context>\` (or optional \`describe\`/\`feature\`), nesting \`when\` -> \`then\` as needed, collapsing single-outcome branches to \`when <action>, then <outcome>\`, and hoisting shared fixtures. Prioritize \`e2e\` > \`integration\` > \`unit\` and specify human-verifiable outputs (screenshots, replays, file dumps, snapshots, JSON outputs, or debug logs). Label skill checks without a harness as \`manual acceptance\`.]
48384
48384
 
48385
48385
  #### Trade-offs
48386
48386
 
@@ -48662,9 +48662,10 @@ var code_review_default = `# Review: [change or pull request]
48662
48662
 
48663
48663
  - **Revision:** [commit or diff]
48664
48664
  - **Intent:** [stated objective and source]
48665
- - **Reviewer:** [name]
48665
+ - **Reviewer:** [actual model identity]
48666
+ - **Mode:** [normal/adversarial]
48666
48667
 
48667
- [Apply root and applicable nested AGENTS.md rules throughout every section. Repository rules take precedence over generic rubric conventions. Cite violations under the relevant section.]
48668
+ [Apply \`/difflab-writing-style\` and \`/difflab-bdd\`. Use short sentences, active voice, and familiar words throughout. Keep section assessments brief and link to findings instead of repeating them. Keep detailed criterion coverage in rubric.yaml. Apply AGENTS.md and named style guides. Informal conventions are context, not requirements. Flag material antipatterns even when existing code shares them.]
48668
48669
 
48669
48670
  ## Intent Delivery
48670
48671
 
@@ -48672,7 +48673,7 @@ var code_review_default = `# Review: [change or pull request]
48672
48673
 
48673
48674
  ## Readability
48674
48675
 
48675
- [Do names and code structure show what happens? Compare names with behavior, side effects, and callers. Follow language conventions and established codebase patterns. Flag misleading names and flows that are hard to follow.]
48676
+ [Do names and code structure show what happens? Compare names with behavior, side effects, and callers. Apply written naming rules. Do not enforce harmless informal patterns. Flag misleading names and flows that are hard to follow.]
48676
48677
 
48677
48678
  ## Semantic Organization
48678
48679
 
@@ -48680,15 +48681,15 @@ var code_review_default = `# Review: [change or pull request]
48680
48681
 
48681
48682
  ## Maintainability
48682
48683
 
48683
- [Does the change follow established codebase patterns and applicable AGENTS.md rules? Check new fields, types, helpers, and dependencies against existing code and callers. Reuse suitable code and remove needless complexity.]
48684
+ [Does the change remain understandable and maintainable under applicable written rules? Existing antipatterns do not justify new ones. Check new fields, types, helpers, and dependencies against existing code and callers. Reuse suitable code and remove needless complexity.]
48684
48685
 
48685
48686
  ## Reliability/Testing
48686
48687
 
48687
- [Prioritize end-to-end application behavior, then module behavior through public APIs, then focused unit behavior with public APIs first. Flag tests that only verify external dependencies rather than the repository's source behavior or integration contract. Require test names to identify the function or functionality tested and its expected behavior, regardless of grouping style. Name missing observable contracts or regressions; do not demand coverage of every function or a percentage. Check semantic grouping, clear Arrange/Act/Assert blocks, fixtures, and existing shared helpers. Flag walls of repeated setup or interleaved actions/assertions that hide the behavior. Report executed checks and limits honestly.]
48688
+ [Apply \`/difflab-bdd\`. Analyze tests against the component's behavior flowchart (converging state branches, \`e2e\` > \`integration\` > \`unit\`, and human-verifiable outputs such as screenshots, replays, file dumps, snapshots, JSON outputs, or debug logs). Propose a concrete \`[M/A/D/R]\` BDD test tree reorganization (\`given\` -> \`when\` -> \`then\` with hoisted fixtures) when tests are flat, repeat downstream assertions across converged states, or obscure setup. Flag tests that only verify external dependencies rather than the repository's source behavior. Report executed checks and limits honestly.]
48688
48689
 
48689
48690
  ## Documentation
48690
48691
 
48691
- [Flag walls of text, complex language, and text that only restates the implementation. Require concise, simple-English explanations. Prefer diagrams when they clarify structures, flows, or relationships better than long prose.]
48692
+ [Apply \`/difflab-writing-style\`. Flag walls of text, complex language, and text that only restates the implementation. Require concise \`/simple-english\` explanations, domain terminology, and Mermaid/Draw.io diagrams or \`[M/A/D/R]\` trees when visuals communicate structure or flow better than prose.]
48692
48693
 
48693
48694
  ## Performance
48694
48695
 
@@ -48696,11 +48697,11 @@ var code_review_default = `# Review: [change or pull request]
48696
48697
 
48697
48698
  ## Security
48698
48699
 
48699
- [Optional: run this dimension only when explicitly requested, including a security focus. Otherwise state that it was not selected. Report clear security defects discovered by other dimensions without claiming a security audit. Avoid unrelated hardening requests.]
48700
+ [Optional in normal mode: select for a security focus or --adversarial. Adversarial mode challenges all applicable dimensions. Otherwise state that it was not selected. Report clear security defects discovered by other dimensions without claiming a security audit. Avoid unrelated hardening requests.]
48700
48701
 
48701
48702
  ## Change Requests
48702
48703
 
48703
- Every generated comment and reply must use simple English and be as short as possible. Include a suggested change when a concrete fix is available. State the issue and its effect directly. Phrase recommendations with \u201CLet's\u201D, \u201CI suggest\u201D, or \u201CI recommend\u201D. Include a short snippet or suggested patch only when it clarifies the fix. Use the scope that fixes the problem, including a refactor when needed. Group repeated instances of one cause. Omit repeated context, nitpicks, minor optimizations, and speculative findings. End each generated request and response with its model attribution quote. Preserve human text unchanged.
48704
+ State each finding once here, with its issue, effect, and useful fix. Use short, plain-English sentences for the whole report, including generated comments and replies. Include a suggested change when a concrete fix is available. State the issue and its effect directly. Phrase recommendations with \u201CLet's\u201D, \u201CI suggest\u201D, or \u201CI recommend\u201D. Include a short snippet or suggested patch only when it clarifies the fix. Use the scope that fixes the problem, including a refactor when needed. Group repeated instances of one cause. Omit repeated context, nitpicks, minor optimizations, and speculative findings. End each generated request and response with its model attribution quote. Preserve human text unchanged.
48704
48705
 
48705
48706
  ### [file:line \u2014 request]
48706
48707
 
@@ -48717,7 +48718,7 @@ Every generated comment and reply must use simple English and be as short as pos
48717
48718
 
48718
48719
  | Severity | Location | Evidence and impact | Required change or question |
48719
48720
  | --- | --- | --- | --- |
48720
- | [blocking/consider/note] | [file:line] | [reproducible observation] | [concrete action] |
48721
+ | [blocking/consider/note] | [file:line] | [short link to Change Request; do not repeat its text] | [short action or link] |
48721
48722
 
48722
48723
  ## Overall Assessment
48723
48724
 
@@ -48733,16 +48734,15 @@ var init_code_review = () => {};
48733
48734
 
48734
48735
  // templates/code-review-rubric.yaml
48735
48736
  var code_review_rubric_default = `version: 1
48736
- worker_defaults:
48737
- model: sol
48738
- reasoning_effort: medium
48737
+ review_defaults:
48738
+ model_policy: retain-frontier
48739
48739
  review:
48740
48740
  base_revision: null
48741
48741
  head_revision: null
48742
48742
  working_tree: null
48743
48743
  changed_files: []
48744
48744
  exclusions: []
48745
- section_reviewers: []
48745
+ reviewers: []
48746
48746
  sections:
48747
48747
  - title: Intent Delivery
48748
48748
  criteria:
@@ -48762,7 +48762,7 @@ sections:
48762
48762
  scope: file
48763
48763
  default_severity: consider
48764
48764
  issue: >-
48765
- Function or other names misrepresent behavior, hide side effects, or break established naming conventions.
48765
+ Function or other names misrepresent behavior, hide side effects, or violate written naming rules.
48766
48766
  results: []
48767
48767
  - id: readability.structure
48768
48768
  scope: file
@@ -48779,7 +48779,7 @@ sections:
48779
48779
  - id: organization.architecture
48780
48780
  scope: review
48781
48781
  default_severity: consider
48782
- issue: Files or classes break the established architecture or module patterns.
48782
+ issue: Files or classes mix responsibilities, obscure ownership, or violate written architecture rules.
48783
48783
  results: []
48784
48784
  - id: organization.api-ergonomics
48785
48785
  scope: applicable-file
@@ -48835,14 +48835,15 @@ sections:
48835
48835
  - id: testing.application
48836
48836
  scope: review
48837
48837
  default_severity: consider
48838
- issue: Required behavior or a meaningful regression lacks test evidence.
48838
+ issue: >-
48839
+ Required behavior lacks outside-in test evidence (e2e > integration > unit), or e2e/integration tests lack human-verifiable outputs such as snapshots, test logs, or JSON outputs.
48839
48840
  results: []
48840
48841
  - id: testing.relevance
48841
48842
  scope: applicable-file
48842
48843
  applies_when: Added or modified test files.
48843
48844
  default_severity: consider
48844
48845
  issue: >-
48845
- A test verifies external dependency behavior without establishing behavior of the repository's own source code.
48846
+ Verbose unit tests pad coverage without proving real contracts, validate static asset or template structure, test the test apparatus itself, add regression tests for changes within the same changeset or for alpha projects without users, or only verify external dependency behavior.
48846
48847
  results: []
48847
48848
  - id: testing.names
48848
48849
  scope: applicable-file
@@ -48856,7 +48857,7 @@ sections:
48856
48857
  applies_when: Added or modified test files.
48857
48858
  default_severity: consider
48858
48859
  issue: >-
48859
- Test grouping or interleaved setup, actions, and assertions hide the scenario or expected result.
48860
+ Test organization lacks clear BDD/Gherkin structure (given/when/then) or interleaves setup, actions, and assertions so coverage or expected results are hard to see.
48860
48861
  results: []
48861
48862
  - id: testing.fixtures
48862
48863
  scope: applicable-file
@@ -48900,8 +48901,6 @@ sections:
48900
48901
  default_severity: blocking
48901
48902
  issue: The change exposes secrets or breaks authorization or a trust boundary.
48902
48903
  results: []
48903
- worker:
48904
- reasoning_effort: high
48905
48904
  `;
48906
48905
  var init_code_review_rubric = () => {};
48907
48906
 
@@ -49245,6 +49244,46 @@ The \`base_commit\` is the full SHA of \`origin/main\` when this branch was crea
49245
49244
  `;
49246
49245
  var init_poc_readme = () => {};
49247
49246
 
49247
+ // templates/hook-definition.md
49248
+ var hook_definition_default = `# {{name}}
49249
+
49250
+ ## Trigger
49251
+
49252
+ {{trigger}}
49253
+
49254
+ ## Action
49255
+
49256
+ - Handler: \`{{handler}}\`
49257
+ - Requires: \`{{requires}}\`
49258
+ - Receives: \`{{receives}}\`
49259
+
49260
+ ## Instructions
49261
+
49262
+ {{instructions}}
49263
+ `;
49264
+ var init_hook_definition = () => {};
49265
+
49266
+ // templates/task.md
49267
+ var task_default = `# [Task title]
49268
+
49269
+ ## Objective
49270
+
49271
+ [One-line description of what needs to be done.]
49272
+
49273
+ ## Context
49274
+
49275
+ [Relevant background written using \`/difflab-writing-style\`, kept as terse as possible. Link to relevant architecture docs instead of repeating them. For a bug ticket, describe only the observable issue and impact; do not propose a solution.]
49276
+
49277
+ ## Technical Approach
49278
+
49279
+ _To be completed by the person doing the work._
49280
+
49281
+ ## References
49282
+
49283
+ - [Relevant links to wikis, architecture docs, tickets, external resources, CI logs, or incident portals.]
49284
+ `;
49285
+ var init_task = () => {};
49286
+
49248
49287
  // templates/examples/plan-feature.md
49249
49288
  var plan_feature_default = `# Invite teammates to a project
49250
49289
 
@@ -49266,8 +49305,6 @@ Project owners need to invite teammates without sharing credentials. Recipients
49266
49305
 
49267
49306
  ### Big Ideas
49268
49307
 
49269
- Keep invitation policy in the service and transport concerns in the handler. Store a single-use token hash, project, recipient, and expiry. Use the existing mail queue; no new service is needed.
49270
-
49271
49308
  \`\`\`mermaid
49272
49309
  flowchart LR
49273
49310
  HTTP["HTTP handler"] --> Service["Invitation service<br/>eligibility and acceptance"]
@@ -49275,7 +49312,7 @@ flowchart LR
49275
49312
  Service --> Queue["Existing mail queue"]
49276
49313
  \`\`\`
49277
49314
 
49278
- Acceptance adds membership and consumes the invitation in one transaction. Invalid acceptance leaves both unchanged. Queue delivery after saving, so an email outage cannot lose an invitation.
49315
+ Store only the token hash. Save before queueing delivery. An email outage must not lose the invitation.
49279
49316
 
49280
49317
  \`\`\`mermaid
49281
49318
  flowchart TD
@@ -49305,34 +49342,50 @@ Never log raw tokens. Apply the existing request rate limit to creation; do not
49305
49342
 
49306
49343
  #### Repo Changes
49307
49344
 
49308
- Illustrative paths: \`+\` new, \`~\` changed, \`*\` largest edits. No files are removed.
49345
+ Illustrative paths: \`[M]\` modified, \`[A]\` added, \`[D]\` deleted, \`[R]\` renamed. No files are removed or renamed. \`service.ts\` contains the largest edits.
49309
49346
 
49310
49347
  \`\`\`text
49311
49348
  src/
49312
49349
  \u251C\u2500\u2500 invitations/
49313
- \u2502 \u251C\u2500\u2500 store.ts + persistence and transaction
49314
- \u2502 \u251C\u2500\u2500 service.ts + * invitation policy
49315
- \u2502 \u2514\u2500\u2500 service.spec.ts + public API with real test store
49350
+ \u2502 \u251C\u2500\u2500 [A] store.ts persistence and transaction
49351
+ \u2502 \u251C\u2500\u2500 [A] service.ts invitation policy
49352
+ \u2502 \u2514\u2500\u2500 [A] service.spec.ts public API with real test store
49316
49353
  \u2514\u2500\u2500 http/
49317
- \u251C\u2500\u2500 routes.ts ~ register invitation handlers
49318
- \u251C\u2500\u2500 invitations.ts + transport and authorization
49319
- \u2514\u2500\u2500 invitations.spec.ts + application request scenarios
49354
+ \u251C\u2500\u2500 [M] routes.ts register invitation handlers
49355
+ \u251C\u2500\u2500 [A] invitations.ts transport and authorization
49356
+ \u2514\u2500\u2500 [A] invitations.spec.ts application request scenarios
49320
49357
  \`\`\`
49321
49358
 
49322
49359
  #### Testing
49323
49360
 
49361
+ \`\`\`mermaid
49362
+ flowchart TD
49363
+ Invite["Owner creates invitation"] --> Saved["State: Pending invitation saved<br/>(even if email queue fails)"]
49364
+ Saved --> Accept["Recipient submits token"]
49365
+ Accept --> Eligible{"Intended recipient,<br/>unused & unexpired?"}
49366
+ Eligible -->|"no"| Reject["Reject; membership & token unchanged"]
49367
+ Eligible -->|"yes"| Tx["Consume token + add member"]
49368
+ Tx -->|"tx fails"| Rollback["Roll back membership & token"]
49369
+ Tx -->|"tx succeeds"| Member["Return active membership"]
49370
+ \`\`\`
49371
+
49324
49372
  \`\`\`text
49325
- end-to-end/
49326
- \u2514\u2500\u2500 invitation acceptance
49327
- \u251C\u2500\u2500 owner sends invite -> recipient accepts -> membership appears
49328
- \u2514\u2500\u2500 wrong recipient, expired, or used token -> rejection; state unchanged
49329
- integration/
49330
- \u2514\u2500\u2500 invitation service
49331
- \u251C\u2500\u2500 failed transaction -> membership and token changes roll back
49332
- \u2514\u2500\u2500 email failure -> saved invitation stays pending for retry
49373
+ src/
49374
+ \u251C\u2500\u2500 http/
49375
+ \u2502 \u2514\u2500\u2500 [A] invitations.spec.ts (application behavior)
49376
+ \u2502 \u2514\u2500\u2500 [A] given a pending owner invitation
49377
+ \u2502 \u251C\u2500\u2500 [A] when the intended recipient accepts, then membership appears in JSON response
49378
+ \u2502 \u2514\u2500\u2500 [A] when a wrong recipient or expired/used token is submitted, then it rejects without changing state
49379
+ \u2514\u2500\u2500 invitations/
49380
+ \u2514\u2500\u2500 [A] service.spec.ts (public module behavior)
49381
+ \u251C\u2500\u2500 [A] given an email queue failure, when creating an invitation, then the saved invitation stays pending for retry
49382
+ \u2514\u2500\u2500 [A] given a pending invitation (converged state)
49383
+ \u2514\u2500\u2500 [A] when the database transaction fails during acceptance
49384
+ \u251C\u2500\u2500 [A] then membership changes roll back
49385
+ \u2514\u2500\u2500 [A] then the invitation token remains unconsumed
49333
49386
  \`\`\`
49334
49387
 
49335
- Use request scenarios in \`src/http/invitations.spec.ts\` and service scenarios in \`src/invitations/service.spec.ts\`. Reuse fixtures with a real test store.
49388
+ Use BDD-style nested \`given\`/\`when\`/\`then\` scenarios in \`src/http/invitations.spec.ts\` and \`src/invitations/service.spec.ts\`. Reuse fixtures with a real test store.
49336
49389
 
49337
49390
  #### Trade-offs
49338
49391
 
@@ -49355,7 +49408,7 @@ Email delivery can fail after persistence. Use existing queue retries and keep t
49355
49408
  - [ ] Queue delivery after persistence, retry transient mail failures, and expose the pending state to owners.
49356
49409
  - [ ] In \`src/http/invitations.spec.ts\`, send an owner invitation, accept as its recipient, and read back membership. Verify wrong-recipient, expired, and used-token requests leave state unchanged. Verify an email failure keeps the saved invitation pending.
49357
49410
  - [ ] Validation
49358
- - [ ] Run the end-to-end acceptance/rejection scenarios, then the service integration scenarios listed above.
49411
+ - [ ] Run the application acceptance/rejection scenarios, then the service integration scenarios listed above.
49359
49412
  - [ ] Run relevant lint and format checks. This example grants no commit or push permission.
49360
49413
  - [ ] If execution authorizes commits and expects CI, commit the changes and confirm CI is green for that commit.
49361
49414
 
@@ -49386,8 +49439,6 @@ Readers should retain access to historical activity while the team changes stora
49386
49439
 
49387
49440
  ### Big Ideas
49388
49441
 
49389
- Write to both stores temporarily, backfill older records with stable IDs, compare counts and ordering, then switch readers behind a reversible flag. Keep the legacy reader authoritative until the new store is verified.
49390
-
49391
49442
  \`\`\`mermaid
49392
49443
  flowchart LR
49393
49444
  Writer["Activity writer"] --> Old["Legacy store"]
@@ -49427,34 +49478,45 @@ readSource = legacy | new # reversible; legacy remains default
49427
49478
 
49428
49479
  #### Repo Changes
49429
49480
 
49430
- Illustrative paths: \`+\` new, \`~\` changed, \`*\` largest edits. No files are removed.
49481
+ Illustrative paths: \`[M]\` modified, \`[A]\` added, \`[D]\` deleted, \`[R]\` renamed. No files are removed or renamed. \`writer.ts\` and \`backfill.ts\` contain the largest edits.
49431
49482
 
49432
49483
  \`\`\`text
49433
49484
  infra/
49434
- \u2514\u2500\u2500 activity-store.tf + store, credentials, write-failure metrics
49485
+ \u2514\u2500\u2500 [A] activity-store.tf store, credentials, write-failure metrics
49435
49486
  src/
49436
49487
  \u2514\u2500\u2500 activity/
49437
- \u251C\u2500\u2500 writer.ts ~ * dual writes and retries
49438
- \u251C\u2500\u2500 backfill.ts + * resumable batches and comparison
49439
- \u251C\u2500\u2500 reader.ts ~ reversible read source
49440
- \u251C\u2500\u2500 migration.spec.ts + public API parity and restart scenarios
49441
- \u2514\u2500\u2500 history.spec.ts ~ application history across cutover/rollback
49488
+ \u251C\u2500\u2500 [M] writer.ts dual writes and retries
49489
+ \u251C\u2500\u2500 [A] backfill.ts resumable batches and comparison
49490
+ \u251C\u2500\u2500 [M] reader.ts reversible read source
49491
+ \u251C\u2500\u2500 [A] migration.spec.ts public API parity and restart scenarios
49492
+ \u2514\u2500\u2500 [M] history.spec.ts application history across cutover/rollback
49442
49493
  \`\`\`
49443
49494
 
49444
49495
  #### Testing
49445
49496
 
49497
+ \`\`\`mermaid
49498
+ flowchart TD
49499
+ Write["writeActivity(event)"] --> Dual["Dual write with stable ID"]
49500
+ Dual -->|"destination fails"| Retry["Retry failed store"] --> Parity["State: Both stores hold identical events"]
49501
+ Backfill["Interrupted backfill"] --> Resume["Resume batch"] --> Parity
49502
+ Parity --> Cutover["Switch read flag to new store"]
49503
+ Cutover --> ReadCutover["Page history across cutover"]
49504
+ Cutover --> Rollback["Write post-cutover events, then roll back flag"]
49505
+ \`\`\`
49506
+
49446
49507
  \`\`\`text
49447
- end-to-end/
49448
- \u2514\u2500\u2500 activity history
49449
- \u251C\u2500\u2500 write and page history across cutover -> same records and cursors
49450
- \u2514\u2500\u2500 write after cutover, then roll back -> no lost events
49451
- integration/
49452
- \u2514\u2500\u2500 activity storage
49453
- \u251C\u2500\u2500 retry a failed destination -> one logical event in each store
49454
- \u2514\u2500\u2500 interrupt and resume backfill -> same IDs and order; no duplicates
49508
+ src/
49509
+ \u2514\u2500\u2500 activity/
49510
+ \u251C\u2500\u2500 [M] history.spec.ts (application behavior)
49511
+ \u2502 \u2514\u2500\u2500 [A] given dual-written stores at parity (converged state)
49512
+ \u2502 \u251C\u2500\u2500 [A] when paging history across reader cutover, then records and cursors match
49513
+ \u2502 \u2514\u2500\u2500 [A] when rolling back the reader after new writes, then no events are lost
49514
+ \u2514\u2500\u2500 [A] migration.spec.ts (public module behavior)
49515
+ \u251C\u2500\u2500 [A] given a failed destination write, when retried, then one logical event exists in each store
49516
+ \u2514\u2500\u2500 [A] given an interrupted backfill batch, when resumed, then IDs and order match without duplicates
49455
49517
  \`\`\`
49456
49518
 
49457
- Use \`src/activity/history.spec.ts\` for application scenarios and \`src/activity/migration.spec.ts\` for public storage APIs with real test stores.
49519
+ Use BDD-style nested \`given\`/\`when\`/\`then\` scenarios in \`src/activity/history.spec.ts\` for application behavior and \`src/activity/migration.spec.ts\` for public storage APIs with real test stores.
49458
49520
 
49459
49521
  #### Trade-offs
49460
49522
 
@@ -49486,7 +49548,7 @@ Dual writes increase latency and allow partial failure. Record discrepancies and
49486
49548
  - [ ] Extend application scenarios in \`src/activity/history.spec.ts\`: write and read paged history, switch stores, write again, and roll back. Assert the same visible history without lost events; do not assert private helper calls.
49487
49549
  - [ ] Keep dual writes and monitor parity through the rollback window.
49488
49550
  - [ ] Validation
49489
- - [ ] Run the end-to-end cutover/rollback scenarios, then the storage integration scenarios listed above.
49551
+ - [ ] Run the application cutover/rollback scenarios, then the storage integration scenarios listed above.
49490
49552
  - [ ] Run relevant lint and format checks. This example grants no commit or push permission.
49491
49553
  - [ ] If execution authorizes commits and expects CI, commit the changes and confirm CI is green for that commit.
49492
49554
 
@@ -49676,94 +49738,74 @@ var init_prd_onboarding = () => {};
49676
49738
  var review_change_default = `# Review: invitation acceptance
49677
49739
 
49678
49740
  - **Revision:** illustrative-change-17
49679
- - **Intent:** Accept a valid invitation and reject an expired one without changing membership.
49741
+ - **Intent:** Accept valid invitations without changing membership on rejection.
49680
49742
  - **Reviewer:** Authored example, not an executed review.
49681
-
49682
- This is a fictional example. A real review must use its runtime model name in attribution.
49743
+ - **Mode:** Normal, single model.
49683
49744
 
49684
49745
  ## Intent Delivery
49685
49746
 
49686
- The change adds an acceptance path. Helper tests do not establish that the request saves membership or leaves it unchanged on rejection.
49747
+ Acceptance lacks a saved-membership check. See [request 2](#request-2).
49687
49748
 
49688
49749
  ## Readability
49689
49750
 
49690
- \`loadInvitation\` consumes the token and creates membership. I suggest \`acceptInvitation\` so callers can see the write.
49751
+ A query-like name hides a write. See [request 1](#request-1).
49691
49752
 
49692
49753
  ## Semantic Organization
49693
49754
 
49694
- The handler maps HTTP to the service. The store owns persistence. No refactor is needed here.
49755
+ The handler maps requests. The service owns acceptance. The store saves membership.
49695
49756
 
49696
49757
  ## Maintainability
49697
49758
 
49698
- The change follows the existing store pattern and reuses invitation status.
49759
+ The service reuses invitation status without adding duplicate state.
49699
49760
 
49700
49761
  ## Reliability/Testing
49701
49762
 
49702
- Helper tests do not prove that acceptance saves membership. Let's add application scenarios for acceptance and rejection, then read membership back.
49703
-
49704
- Repeated setup and nested assertions hide the scenario in \`service.spec.ts:18\`. I suggest a fixture and separate blocks:
49763
+ See [request 2](#request-2).
49705
49764
 
49706
- \`\`\`ts
49707
- describe('acceptInvitation', () => {
49708
- test('rejects an expired invitation without adding membership', async () => {
49709
- // Arrange
49710
- const { service, members, recipient, expiredToken } = await invitationFixture()
49711
-
49712
- // Act
49713
- const acceptance = service.acceptInvitation(expiredToken, recipient.id)
49714
-
49715
- // Assert
49716
- await expect(acceptance).rejects.toThrow(InvitationExpiredError)
49717
- expect(await members.list(recipient.id)).toEqual([])
49718
- })
49719
- })
49720
- \`\`\`
49721
-
49722
- The fixture creates a real test store. I recommend keeping the action and assertions visible and reusing shared setup.
49723
-
49724
- ## Repository Guidance
49765
+ ## Documentation
49725
49766
 
49726
- The fictional \`AGENTS.md\` requires Arrange/Act/Assert blocks and local helpers at the bottom. I suggest separating the mixed steps in \`service.spec.ts:18\`.
49767
+ The changed API comment explains token consumption.
49727
49768
 
49728
49769
  ## Performance
49729
49770
 
49730
- SKIPPED: no concrete performance regression is shown and no deep performance audit was requested.
49771
+ No material regression found by inspection. No performance measurements ran.
49731
49772
 
49732
49773
  ## Security
49733
49774
 
49734
- SKIPPED: no concrete security defect is shown; an exhaustive security audit is outside this example's scope.
49775
+ Not selected. No security audit is claimed.
49735
49776
 
49736
49777
  ## Change Requests
49737
49778
 
49738
- ### src/invitations/service.ts:42 \u2014 Name the state-changing operation
49779
+ ### Request 1
49739
49780
 
49740
49781
  - **Status:** open
49741
- - **Request:** \`loadInvitation\` also writes membership. I suggest renaming it to \`acceptInvitation\` and updating callers.
49782
+ - **Request:** \`src/invitations/service.ts:42\`: \`loadInvitation\` also writes membership. I suggest \`acceptInvitation\` so callers see the write.
49742
49783
 
49743
- > Generated via [actual model name from runtime]
49784
+ > Generated via [actual runtime model; illustrative placeholder only]
49744
49785
 
49745
- ### src/invitations/service.spec.ts:18 \u2014 Expose the scenario and test its public result
49786
+ ### Request 2
49746
49787
 
49747
49788
  - **Status:** open
49748
- - **Request:** I suggest the fixture and Arrange/Act/Assert blocks above. Let's add acceptance/rejection scenarios that check saved membership, not just helpers.
49789
+ - **Request:** \`src/invitations/service.spec.ts:18\`: Helper tests do not establish saved membership. Let's test acceptance and rejection through the public API, then read membership back.
49749
49790
 
49750
- > Generated via [actual model name from runtime]
49791
+ > Generated via [actual runtime model; illustrative placeholder only]
49751
49792
 
49752
49793
  ## Findings
49753
49794
 
49754
- | Severity | Location | Evidence and impact | Required change or question |
49755
- | --- | --- | --- | --- |
49756
- | consider | \`src/invitations/service.ts:42\` | A query-like name hides token consumption and membership creation. | I suggest renaming the operation and callers. |
49757
- | consider | \`src/invitations/service.spec.ts:18\` | Repeated setup hides the action; helper coverage does not establish saved membership or rejection behavior. | I suggest clear blocks, a fixture, and a public/application scenario. |
49795
+ | Severity | Location | Request |
49796
+ | -------- | -------------------- | ----------------------- |
49797
+ | consider | \`service.ts:42\` | [Request 1](#request-1) |
49798
+ | consider | \`service.spec.ts:18\` | [Request 2](#request-2) |
49758
49799
 
49759
49800
  ## Overall Assessment
49760
49801
 
49761
- I recommend clearer API naming and acceptance/rejection scenarios. No private-helper coverage or speculative clock-skew request is needed.
49802
+ Address the two requests. No unrelated cleanup is needed.
49762
49803
 
49763
49804
  ## Verification and Limits
49764
49805
 
49765
- - **Checks run:** none; this is an authored example, not a report of passing tests.
49766
- - **Not verified:** execution, actual repository guidance, real email delivery, and production behavior.
49806
+ - **Checks run:** None. This is a fictional example.
49807
+ - **Not verified:** Execution and real repository guidance.
49808
+ - **Rubric:** A real review links its filled rubric here.
49767
49809
  `;
49768
49810
  var init_review_change = () => {};
49769
49811
 
@@ -49935,6 +49977,8 @@ var init_source = __esm(() => {
49935
49977
  init_exploration_research();
49936
49978
  init_exploration_proposal();
49937
49979
  init_poc_readme();
49980
+ init_hook_definition();
49981
+ init_task();
49938
49982
  init_plan_feature();
49939
49983
  init_plan_migration();
49940
49984
  init_design_service();
@@ -49962,6 +50006,8 @@ var init_source = __esm(() => {
49962
50006
  "exploration-research.md": exploration_research_default,
49963
50007
  "exploration-proposal.md": exploration_proposal_default,
49964
50008
  "poc-readme.md": poc_readme_default,
50009
+ "hook-definition.md": hook_definition_default,
50010
+ "task.md": task_default,
49965
50011
  "examples/plan-feature.md": plan_feature_default,
49966
50012
  "examples/plan-migration.md": plan_migration_default,
49967
50013
  "examples/design-service.md": design_service_default,
@@ -49993,7 +50039,9 @@ var init_source = __esm(() => {
49993
50039
  { name: "exploration-summary", description: "Source-grounded technical exploration summary" },
49994
50040
  { name: "exploration-research", description: "Technical exploration research notes" },
49995
50041
  { name: "exploration-proposal", description: "Optional technical exploration approach" },
49996
- { name: "poc-readme", description: "Disposable proof-of-concept branch README" }
50042
+ { name: "poc-readme", description: "Disposable proof-of-concept branch README" },
50043
+ { name: "hook-definition", description: "Reusable global Agent Skill hook definition" },
50044
+ { name: "task", description: "Terse task or bug ticket with objective, context, and references" }
49997
50045
  ];
49998
50046
  embeddedSource = {
49999
50047
  catalog,
@@ -52552,7 +52600,7 @@ function createProgram(write = console.log, version = "0.1.0", options = {}) {
52552
52600
 
52553
52601
  // src/index.ts
52554
52602
  init_errors3();
52555
- var version2 = "0.11.0";
52603
+ var version2 = "0.13.0";
52556
52604
  try {
52557
52605
  await createProgram(console.log, version2).parseAsync(process.argv);
52558
52606
  } catch (error) {
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@difflab/difflab-cli",
3
- "version": "0.11.0",
3
+ "version": "0.13.0",
4
4
  "private": false,
5
5
  "description": "A minimal Bun CLI starter built with Commander",
6
6
  "license": "Apache-2.0",