@nforma.ai/nforma 0.2.1 → 0.29.0

package/commands/{qgsd → nf}/solve.md

@@ -1,5 +1,5 @@
 ---
-name: qgsd:solve
+name: nf:solve
 description: Orchestrator skill that migrates legacy .formal/ layouts, diagnoses consistency gaps, dispatches to remediation skills for each gap type, and converges via diagnose-remediate-rediagnose loop with before/after comparison
 argument-hint: [--report-only] [--max-iterations=N] [--json] [--verbose]
 allowed-tools:
@@ -14,7 +14,7 @@ allowed-tools:
 ---
 
 <objective>
-Run the QGSD consistency solver as a full orchestrator. Sweeps 7 layer transitions (R->F, F->T, C->F, T->C, F->C, R->D, D->C), computes a residual vector showing gaps at each boundary, and automatically dispatches to the correct remediation skill/script for each gap type. Re-diagnoses after each remediation round and iterates until convergence or max iterations reached. Returns before/after residual comparison.
+Run the nForma consistency solver as a full orchestrator. Sweeps 7 layer transitions (R->F, F->T, C->F, T->C, F->C, R->D, D->C), computes a residual vector showing gaps at each boundary, and automatically dispatches to the correct remediation skill/script for each gap type. Re-diagnoses after each remediation round and iterates until convergence or max iterations reached. Returns before/after residual comparison.
 </objective>
 
 <execution_context>
@@ -23,11 +23,11 @@ any questions. Do NOT stop for human input. If a sub-skill fails, log the
 failure and continue to the next gap. The only valid reason to stop is:
 all iterations exhausted, or total residual is zero.
 
-This is a self-contained orchestrator skill. It runs the diagnostic engine (bin/qgsd-solve.cjs) and orchestrates higher-level remediation via sub-skills and scripts. No external quorum dispatch is needed — quorum enforcement, if required, is the responsibility of the sub-skills being called.
+This is a self-contained orchestrator skill. It runs the diagnostic engine (bin/nf-solve.cjs) and orchestrates higher-level remediation via sub-skills and scripts. No external quorum dispatch is needed — quorum enforcement, if required, is the responsibility of the sub-skills being called.
 
 BULK REMEDIATION: For F->T and R->D gaps, the solve skill writes PLAN.md files
-directly and dispatches qgsd-executor agents — it does NOT invoke
-/qgsd:quick for bulk remediation. This avoids per-batch quorum overhead while
+directly and dispatches nf-executor agents — it does NOT invoke
+/nf:quick for bulk remediation. This avoids per-batch quorum overhead while
 maintaining quality through the convergence loop's before/after verification.
 The solve skill IS the planner for these mechanical remediation tasks.
 
@@ -46,10 +46,10 @@ Before running the diagnostic sweep, check for a legacy `.formal/` directory at
 Run the migration script using absolute paths (or fall back to CWD-relative):
 
 ```bash
-MIGRATE=$(node ~/.claude/qgsd-bin/migrate-formal-dir.cjs --json --project-root=$(pwd) 2>&1)
+MIGRATE=$(node ~/.claude/nf-bin/migrate-formal-dir.cjs --json --project-root=$(pwd) 2>&1)
 ```
 
-If `~/.claude/qgsd-bin/migrate-formal-dir.cjs` does not exist, fall back to `bin/migrate-formal-dir.cjs` (CWD-relative).
+If `~/.claude/nf-bin/migrate-formal-dir.cjs` does not exist, fall back to `bin/migrate-formal-dir.cjs` (CWD-relative).
 If neither exists, skip this step silently — the migration script is optional for projects that never had a legacy layout.
 
 Parse the JSON output:
@@ -63,11 +63,11 @@ Parse the JSON output:
 Run the diagnostic solver using absolute paths (or fall back to CWD-relative):
 
 ```bash
-BASELINE=$(node ~/.claude/qgsd-bin/qgsd-solve.cjs --json --report-only --project-root=$(pwd))
+BASELINE=$(node ~/.claude/nf-bin/nf-solve.cjs --json --report-only --project-root=$(pwd))
 ```
 
-If ~/.claude/qgsd-bin/qgsd-solve.cjs does not exist, fall back to bin/qgsd-solve.cjs (CWD-relative).
-If neither exists, error with: "QGSD solve scripts not installed. Run `node bin/install.js --claude --global` from the QGSD repo."
+If ~/.claude/nf-bin/nf-solve.cjs does not exist, fall back to bin/nf-solve.cjs (CWD-relative).
+If neither exists, error with: "nForma solve scripts not installed. Run `node bin/install.js --claude --global` from the nForma repo."
 
 Parse the JSON output to extract the `residual_vector` object. Key fields:
 - `residual_vector.r_to_f.residual` — count of requirements lacking formal coverage
@@ -118,12 +118,12 @@ Extract the list of uncovered requirement IDs from `residual_vector.r_to_f.detai
 
 If the list has 10 or fewer IDs, dispatch:
 ```
-/qgsd:close-formal-gaps --batch --ids=REQ-01,REQ-02,...
+/nf:close-formal-gaps --batch --ids=REQ-01,REQ-02,...
 ```
 
 If the list has more than 10 IDs, dispatch:
 ```
-/qgsd:close-formal-gaps --batch --all
+/nf:close-formal-gaps --batch --all
 ```
 
 Log: `"Dispatching R->F remediation: close-formal-gaps for {N} uncovered requirements"`
@@ -142,10 +142,10 @@ Find tool JARs at: `.planning/formal/tla/tla2tools.jar` (or `~/.claude/.planning
 
 **Phase 1 — Generate stubs:** Run the formal-test-sync script to generate test stubs and update traceability sidecars:
 ```bash
-node ~/.claude/qgsd-bin/formal-test-sync.cjs --project-root=$(pwd)
+node ~/.claude/nf-bin/formal-test-sync.cjs --project-root=$(pwd)
 ```
 
-If ~/.claude/qgsd-bin/formal-test-sync.cjs does not exist, fall back to bin/formal-test-sync.cjs (CWD-relative).
+If ~/.claude/nf-bin/formal-test-sync.cjs does not exist, fall back to bin/formal-test-sync.cjs (CWD-relative).
 
 Log: `"F->T phase 1: formal-test-sync generated {N} stubs"`
 
@@ -164,7 +164,7 @@ console.log('[solve] Recipes: ' + recipes.length + ' total, ' + incomplete.lengt
 ```
 Incomplete recipes (missing source_files or definition) produce lower-quality tests but do NOT block dispatch.
 
-**Phase 2 — Implement stubs via direct parallel executor dispatch:** Stubs alone do not close the gap — they contain `assert.fail('TODO')`. The solver dispatches `qgsd-executor` agents directly to implement real test logic — it does NOT use `/qgsd:quick` for bulk stub implementation.
+**Phase 2 — Implement stubs via direct parallel executor dispatch:** Stubs alone do not close the gap — they contain `assert.fail('TODO')`. The solver dispatches `nf-executor` agents directly to implement real test logic — it does NOT use `/nf:quick` for bulk stub implementation.
 
 1. **Load context:** Parse `.planning/formal/formal-test-sync-report.json` for each stub's `requirement_id`, `formal_properties[].model_file`, `formal_properties[].property`. Also verify recipe files exist at `.planning/formal/generated-stubs/{ID}.stub.recipe.json` — these contain pre-resolved context (requirement text, property definition, source files, import hints, test strategy).
 
@@ -229,9 +229,9 @@ Incomplete recipes (missing source_files or definition) produce lower-quality te
 
 4. **Spawn executors in sequential waves of 3** — To avoid OOM on developer machines (each executor consumes ~1GB RAM), dispatch at most 3 parallel executors at a time. Wait for each wave to finish before starting the next:
    ```
-   Wave 1: Task(subagent_type="qgsd-executor", description="F->T stubs batch 1"), batch 2, batch 3
+   Wave 1: Task(subagent_type="nf-executor", description="F->T stubs batch 1"), batch 2, batch 3
    [wait for all 3 to complete]
-   Wave 2: Task(subagent_type="qgsd-executor", description="F->T stubs batch 4"), batch 5, batch 6
+   Wave 2: Task(subagent_type="nf-executor", description="F->T stubs batch 4"), batch 5, batch 6
    [wait for all 3 to complete]
    ... continue until all batches dispatched
    ```
@@ -250,7 +250,7 @@ The T->C residual counts both failures and skipped tests. Extract detail:
 
 Dispatch the fix-tests skill:
 ```
-/qgsd:fix-tests
+/nf:fix-tests
 ```
 
 This will discover and autonomously fix failing AND skipped tests. Skipped tests often indicate incomplete implementations or platform-specific guards that need resolution.
@@ -261,7 +261,7 @@ If it fails, log the failure and continue.
 
 ### 3d. C->F Gaps (residual_vector.c_to_f.residual > 0)
 
-Constant mismatches between code and formal specs. Display the mismatch table, then dispatch `/qgsd:quick` to align them:
+Constant mismatches between code and formal specs. Display the mismatch table, then dispatch `/nf:quick` to align them:
 
 ```
 C->F: {N} constant mismatch(es) — dispatching quick task to align
@@ -276,7 +276,7 @@ constant_name   formal_spec_file  formal_val      config_val
 
 Dispatch:
 ```
-/qgsd:quick Fix C->F constant mismatches: update formal specs OR code config to align these values: {mismatch_summary}
+/nf:quick Fix C->F constant mismatches: update formal specs OR code config to align these values: {mismatch_summary}
 ```
 
 If the mismatch has `intentional_divergence: true`, skip it and log as intentional.
@@ -285,19 +285,19 @@ If the mismatch has `intentional_divergence: true`, skip it and log as intention
 
 First, run the formal verification using absolute paths to get fresh failure data:
 ```bash
-node ~/.claude/qgsd-bin/run-formal-verify.cjs --project-root=$(pwd)
+node ~/.claude/nf-bin/run-formal-verify.cjs --project-root=$(pwd)
 ```
 
-If ~/.claude/qgsd-bin/run-formal-verify.cjs does not exist, fall back to bin/run-formal-verify.cjs (CWD-relative).
+If ~/.claude/nf-bin/run-formal-verify.cjs does not exist, fall back to bin/run-formal-verify.cjs (CWD-relative).
 
 Then parse `.planning/formal/check-results.ndjson` and classify each failure:
 
 | Classification | Criteria | Dispatch |
 |---------------|----------|----------|
-| **Syntax error** | Summary contains "Syntax error", "parse error" | `/qgsd:quick Fix Alloy/TLA+ syntax error in {model_file}: {error_detail}` |
-| **Scope error** | Summary contains "scope", "sig" | `/qgsd:quick Fix scope declaration in {model_file}: {error_detail}` |
-| **Conformance divergence** | check_id contains "conformance" | `/qgsd:quick Fix conformance trace divergences in {model_file}: {error_detail}` |
-| **Verification failure** | Counterexample found | `/qgsd:quick Fix formal verification counterexample in {check_id}: {summary}` |
+| **Syntax error** | Summary contains "Syntax error", "parse error" | `/nf:quick Fix Alloy/TLA+ syntax error in {model_file}: {error_detail}` |
+| **Scope error** | Summary contains "scope", "sig" | `/nf:quick Fix scope declaration in {model_file}: {error_detail}` |
+| **Conformance divergence** | check_id contains "conformance" | `/nf:quick Fix conformance trace divergences in {model_file}: {error_detail}` |
+| **Verification failure** | Counterexample found | `/nf:quick Fix formal verification counterexample in {check_id}: {summary}` |
 | **Missing tool** | "not found", "not installed" | Log as infrastructure gap, skip |
 | **Inconclusive** | result = "inconclusive" | Skip — not a failure |
 
@@ -321,7 +321,7 @@ R->D: {N} requirement(s) undocumented in developer docs:
   ...
 ```
 
-Then auto-remediate by dispatching a single `qgsd-executor` agent directly — it does NOT use `/qgsd:quick` for bulk doc generation.
+Then auto-remediate by dispatching a single `nf-executor` agent directly — it does NOT use `/nf:quick` for bulk doc generation.
 
 1. Read `.planning/formal/requirements.json` to get the text/description for each undocumented requirement ID.
 2. For each undocumented ID, identify the most relevant source file(s) by grepping the codebase for the requirement ID and its key terms (use Grep tool).
@@ -342,7 +342,7 @@ Then auto-remediate by dispatching a single `qgsd-executor` agent directly — i
 
 4. **Spawn ONE executor:**
    ```
-   Task(subagent_type="qgsd-executor", description="R->D: generate doc entries for {N} requirements")
+   Task(subagent_type="nf-executor", description="R->D: generate doc entries for {N} requirements")
    ```
    Wait for it to complete. If it fails, log the failure and continue.
 
@@ -406,7 +406,7 @@ Accept: [a]ll / [n]one / comma-separated numbers (e.g. 1,3,5) / [s]kip this cycl
 
 Wait for user input via AskUserQuestion. Route based on response:
 
-- **Numbers or "all"**: For each accepted candidate, dispatch `/qgsd:add-requirement` with the candidate evidence as context. The add-requirement skill handles ID assignment, duplicate checks, and semantic conflict detection. Approved candidates enter the forward flow (R→F→T→C) in the next iteration.
+- **Numbers or "all"**: For each accepted candidate, dispatch `/nf:add-requirement` with the candidate evidence as context. The add-requirement skill handles ID assignment, duplicate checks, and semantic conflict detection. Approved candidates enter the forward flow (R→F→T→C) in the next iteration.
 
 - **"none"**: Write ALL candidates to `.planning/formal/acknowledged-not-required.json` so they are not resurfaced in future runs. Each entry:
   ```json
@@ -426,10 +426,10 @@ Log: `"Reverse discovery: {N} candidates presented, {M} approved, {K} rejected,
 
 After all remediations in Step 3 complete, run the diagnostic again using absolute paths:
 ```bash
-POST=$(node ~/.claude/qgsd-bin/qgsd-solve.cjs --json --report-only --project-root=$(pwd))
+POST=$(node ~/.claude/nf-bin/nf-solve.cjs --json --report-only --project-root=$(pwd))
 ```
 
-If ~/.claude/qgsd-bin/qgsd-solve.cjs does not exist, fall back to bin/qgsd-solve.cjs (CWD-relative).
+If ~/.claude/nf-bin/nf-solve.cjs does not exist, fall back to bin/nf-solve.cjs (CWD-relative).
 
 Parse the result as `post_residual`.
 
@@ -504,10 +504,10 @@ Note: R->D gaps are auto-remediated by generating developer doc entries in docs/
 
 After the before/after table, run the full formal verification using absolute paths if not already run during Step 3e:
 ```bash
-node ~/.claude/qgsd-bin/run-formal-verify.cjs --project-root=$(pwd)
+node ~/.claude/nf-bin/run-formal-verify.cjs --project-root=$(pwd)
 ```
 
-If ~/.claude/qgsd-bin/run-formal-verify.cjs does not exist, fall back to bin/run-formal-verify.cjs (CWD-relative).
+If ~/.claude/nf-bin/run-formal-verify.cjs does not exist, fall back to bin/run-formal-verify.cjs (CWD-relative).
 
 Parse `.planning/formal/check-results.ndjson` and display **every check** grouped by result:
 
@@ -535,7 +535,7 @@ Display checks in this order: PASS first (alphabetical), then FAIL (alphabetical
 After the table, if there are any FAIL or INCONCLUSIVE checks, add a brief actionability note:
 ```
 {fail_count} check(s) failing, {inconc_count} inconclusive.
-Failing checks need investigation — use /qgsd:quick to dispatch fixes for syntax/scope errors or conformance divergences.
+Failing checks need investigation — use /nf:quick to dispatch fixes for syntax/scope errors or conformance divergences.
 Inconclusive checks are not failures but indicate incomplete verification (usually missing fairness declarations or tools).
 ```
 
@@ -543,7 +543,7 @@ This table is mandatory even when the solver layer residuals are all zero — be
 
 ## Important Constraints
 
-1. **bin/qgsd-solve.cjs is NOT modified** — it remains the diagnostic engine. This skill orchestrates remediation at the skill/script level.
+1. **bin/nf-solve.cjs is NOT modified** — it remains the diagnostic engine. This skill orchestrates remediation at the skill/script level.
 
 2. **Convergence loop is at skill level** — when the skill calls diagnostic again in Step 4, it uses `--json --report-only` to get fresh data. The skill then decides whether to loop back to Step 3 or exit. The script's internal auto-close loop is bypassed.
 

package/commands/{qgsd → nf}/sync-baselines.md

@@ -1,5 +1,5 @@
 ---
-name: qgsd:sync-baselines
+name: nf:sync-baselines
 description: Sync baseline requirements into .planning/formal/requirements.json (auto-detects project intent by default)
 argument-hint: [--profile <web|mobile|desktop|api|cli|library>]
 allowed-tools:
@@ -9,7 +9,7 @@ allowed-tools:
 ---
 
 <objective>
-Sync baseline requirements from the QGSD defaults into `.planning/formal/requirements.json`. Auto-detects project intent by default by scanning the repo for framework, deployment, and configuration signals. Supports explicit `--profile` override. Runs `node bin/sync-baseline-requirements.cjs`, displays results, and commits if requirements were added.
+Sync baseline requirements from the nForma defaults into `.planning/formal/requirements.json`. Auto-detects project intent by default by scanning the repo for framework, deployment, and configuration signals. Supports explicit `--profile` override. Runs `node bin/sync-baseline-requirements.cjs`, displays results, and commits if requirements were added.
 </objective>
 
 <process>
@@ -97,13 +97,13 @@ fs.writeFileSync(configPath, JSON.stringify(config, null, 2) + '\n');
 If `added.length > 0`:
 
 ```bash
-node ~/.claude/qgsd/bin/gsd-tools.cjs commit "req(baseline): sync N baseline requirements" --files .planning/formal/requirements.json
+node ~/.claude/nf/bin/gsd-tools.cjs commit "req(baseline): sync N baseline requirements" --files .planning/formal/requirements.json
 ```
 
 Also commit config if intent was stored:
 
 ```bash
-node ~/.claude/qgsd/bin/gsd-tools.cjs commit "chore(baseline): store detected project intent" --files .planning/config.json
+node ~/.claude/nf/bin/gsd-tools.cjs commit "chore(baseline): store detected project intent" --files .planning/config.json
 ```
 
 Where N is the count of added requirements.

package/commands/{qgsd → nf}/triage.md

@@ -1,5 +1,5 @@
 ---
-name: qgsd:triage
+name: nf:triage
 description: Fetch issues/errors from configured sources (GitHub, Sentry, custom) and triage them. Sources are defined in .planning/triage-sources.md
 argument-hint: "[--source github|sentry|sentry-feedback|bash] [--since 24h|7d] [--limit N]"
 allowed-tools:
@@ -11,7 +11,7 @@ allowed-tools:
 ---
 
 <objective>
-Aggregate issues and errors from all configured sources, deduplicate, render a prioritized triage table, and route the selected issue to the right QGSD workflow.
+Aggregate issues and errors from all configured sources, deduplicate, render a prioritized triage table, and route the selected issue to the right nForma workflow.
 
 This command is the project's unified "what's broken right now?" entry point.
 </objective>
@@ -36,7 +36,7 @@ Parse the YAML frontmatter block (between `---` delimiters at the top of the fil
 Display:
 ```
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- QGSD ► TRIAGE: No sources configured
+ nForma ► TRIAGE: No sources configured
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
 Create .planning/triage-sources.md to configure issue sources.
@@ -50,7 +50,7 @@ Stop.
 
 ```
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- QGSD ► TRIAGE: Fetching from N source(s)...
+ nForma ► TRIAGE: Fetching from N source(s)...
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 ```
 (Replace N with actual count.)
@@ -172,7 +172,7 @@ Sort by:
 If total issues = 0:
 ```
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- QGSD ► TRIAGE: All clear — no open issues found
+ nForma ► TRIAGE: All clear — no open issues found
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 Sources checked: <list>
 ```
@@ -181,7 +181,7 @@ Stop.
 Otherwise:
 ```
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
- QGSD ► TRIAGE: N issues across M source(s)
+ nForma ► TRIAGE: N issues across M source(s)
 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
 
 ┌────┬─────────────────────────────────────────────┬──────────────┬─────────┬────────────┐
@@ -208,15 +208,15 @@ Enter issue # to work on, "all" for full details, or press Enter to skip:
 **If user enters a number:**
 - Load the full issue details (title, URL, meta) for that index.
 - Determine routing:
-  - `severity: error` or `severity: bug` → suggest `/qgsd:debug`
-  - `severity: warning` or `severity: info` → suggest `/qgsd:quick`
+  - `severity: error` or `severity: bug` → suggest `/nf:debug`
+  - `severity: warning` or `severity: info` → suggest `/nf:quick`
 - Display:
   ```
   ◆ Issue: <title>
     URL: <url>
     Meta: <meta>
 
-  Suggested action: /qgsd:debug "<title> — <meta>"
+  Suggested action: /nf:debug "<title> — <meta>"
   Run it? [Y/n]
   ```
 - If confirmed, invoke the suggested skill with the issue as context.
@@ -227,7 +227,7 @@ Enter issue # to work on, "all" for full details, or press Enter to skip:
 
 **If user presses Enter (blank):**
 ```
-Triage skipped. Run /qgsd:triage again when ready.
+Triage skipped. Run /nf:triage again when ready.
 ```
 
 </process>

package/commands/{qgsd → nf}/update.md

@@ -1,5 +1,5 @@
 ---
-name: qgsd:update
+name: nf:update
 description: Update GSD to latest version with changelog display
 allowed-tools:
   - Bash
@@ -19,11 +19,11 @@ Routes to the update workflow which handles:
 </objective>
 
 <execution_context>
-@~/.claude/qgsd/workflows/update.md
+@~/.claude/nf/workflows/update.md
 </execution_context>
 
 <process>
-**Follow the update workflow** from `@~/.claude/qgsd/workflows/update.md`.
+**Follow the update workflow** from `@~/.claude/nf/workflows/update.md`.
 
 The workflow handles all logic including:
 1. Installed version detection (local/global)

package/commands/{qgsd → nf}/verify-work.md

@@ -1,5 +1,5 @@
 ---
-name: qgsd:verify-work
+name: nf:verify-work
 description: Validate built features through conversational UAT
 argument-hint: "[phase number, e.g., '4']"
 allowed-tools:
@@ -16,12 +16,12 @@ Validate built features through conversational testing with persistent state.
 
 Purpose: Confirm what Claude built actually works from user's perspective. One test at a time, plain text responses, no interrogation. When issues are found, automatically diagnose, plan fixes, and prepare for execution.
 
-Output: {phase_num}-UAT.md tracking all test results. If issues found: diagnosed gaps, verified fix plans ready for /qgsd:execute-phase
+Output: {phase_num}-UAT.md tracking all test results. If issues found: diagnosed gaps, verified fix plans ready for /nf:execute-phase
 </objective>
 
 <execution_context>
-@~/.claude/qgsd/workflows/verify-work.md
-@~/.claude/qgsd/templates/UAT.md
+@~/.claude/nf/workflows/verify-work.md
+@~/.claude/nf/templates/UAT.md
 </execution_context>
 
 <context>
@@ -33,6 +33,6 @@ Context files are resolved inside the workflow (`init verify-work`) and delegate
 </context>
 
 <process>
-Execute the verify-work workflow from @~/.claude/qgsd/workflows/verify-work.md end-to-end.
+Execute the verify-work workflow from @~/.claude/nf/workflows/verify-work.md end-to-end.
 Preserve all workflow gates (session management, test presentation, diagnosis, fix planning, routing).
 </process>