@iowarp/clio-coder 0.4.9 → 0.5.0

package/docs/process/documentation-coverage.md

@@ -125,3 +125,11 @@ Per-run verification evidence for the refresh lives with the maintenance records
 for that sprint, not in this page. This page tracks documentation coverage, and
 run logs age out of usefulness faster than the docs they checked.
 
+## Ongoing Coverage and Maintenance Guidelines
+
+Documentation in `docs/` is maintained alongside code changes according to the following durable principles:
+
+- **Source anchoring:** Conceptual descriptions and feature summaries cite concrete implementation modules and contracts under `src/` or `tests/`, avoiding speculative capabilities or unbacked UI claims.
+- **Terminal-first framing:** Clio Coder is documented primarily as a terminal coding agent. Early GUI capabilities (`clio-coder gui`, `clio-coder docs`) are presented as developer preview tools for local browser viewing and telemetry inspection, distinguishing active terminal runtime behavior from recovered workbench reference specifications.
+- **Historical distinction:** Dated audit artifacts, incident reports, and design proposals must be clearly distinguished from active operator guides and architecture references via prominent status notices.
+- **Separation of receipts:** Test logs, per-commit audit inventories, and execution receipts belong in local scratch or release evidence outside the documentation tree.

package/docs/process/documentation-guide.md

@@ -43,7 +43,9 @@ Classify claims clearly:
 | [configuration-reference.md](../guide/configuration-reference.md) | `src/core/defaults.ts`, `src/cli/**`, `src/tools/**`, `src/domains/agents/recipe-schema.ts`, `src/domains/providers/models/local-models/**` | Audited inventory of every settings key, environment variable, CLI flag, project-file key, frontmatter key, tool argument, and model tag with defaults and precedence. |
 | [safety-model.md](../architecture/safety-model.md) | `src/domains/safety/**`, `src/tools/registry.ts`, `src/tools/policy.ts`, `src/tools/verify/**`, `src/entry/orchestrator.ts`, `src/domains/dispatch/write-boundary.ts` | Operating posture, `resolveEffectiveAutonomy` / `resolveBaselineAutonomy`, detect-and-rollback write boundaries, approval axes, damage control, typed validation. |
 | [prompt-envelope-and-tools.md](../architecture/prompt-envelope-and-tools.md) | `src/domains/prompts/compiler.ts`, `src/interactive/chat-loop.ts`, `src/core/tool-names.ts`, `src/tools/registry.ts`, `src/tools/observation.ts`, `src/tools/agent-tools.ts` | Prompt envelope reuse, canonical tool delivery via single `agent-tools.ts` adapter, seven-plane tool surface, observation envelope, strict `ToolName` keying. |
-| [tool-usage.md](../guide/tool-usage.md) | `src/tools/agent-tools.ts`, `src/tools/registry.ts`, `src/tools/observation.ts` | In-depth reference for all 24 canonical tools, including conditional registration, parameters, typical payloads, normalizers, and error examples. |
+| [tool-usage.md](../guide/tool-usage.md) | `src/tools/agent-tools.ts`, `src/tools/registry.ts`, `src/tools/observation.ts` | In-depth reference for all 30 built-in tools across 8 planes (21 direct tools, 9 gateway capabilities), including conditional registration, parameters, typical payloads, normalizers, and error examples. |
+| [doctor.md](../guide/doctor.md) | `src/cli/doctor.ts` | Diagnostic checks across install structure, credentials, HPC toolchain, and model targets (`doctor`, `doctor --fix`, `doctor --deep`). |
+| [slurm.md](../guide/slurm.md) | `src/tools/gateway/**`, `library/skills/slurm/**` | Slurm job submission, polling, and cancellation via the `clio-kit` stdio MCP server. |
 | [panes-and-files.md](../guide/panes-and-files.md) | `src/interactive/panes-runtime.ts`, `src/interactive/yazi-bridge.ts`, `src/domains/mux/operations.ts`, `src/domains/mux/yazi/theme.ts`, `src/cli/panes.ts` | Operator path to panes and the files pane: install, settings, doctor rows, `/files` and `Alt+E`, presets, what `/quit` closes, theme boundary, troubleshooting. |
 | [provider-adapter-cookbook.md](../architecture/provider-adapter-cookbook.md) | `src/domains/providers/registry.ts`, `src/domains/providers/types/runtime-descriptor.ts` | RuntimeDescriptor, probe(), probeReasoning(), synthesizeModel(), thinking mechanisms. |
 | [pi-boundary.md](../architecture/pi-boundary.md) | `src/engine/**`, `src/tools/truncate.ts`, `tests/boundaries/check-boundaries.ts` | Which Pi SDK capabilities Clio delegates to, which Clio-specific adapters remain, and the boundary tests that keep the split explicit. |
@@ -88,6 +90,21 @@ Classify claims clearly:
 | [tool-audit-v0.4.9.md](tool-audit-v0.4.9.md) | `src/tools/**`, `src/core/safe-exec.ts`, `src/domains/gateway/mcp/**` | Individual tool purpose, placement, contracts, implementation bounds, scientific integrity, and landed disposition. |
 | [scientific-validation.md](scientific-validation.md) | `src/domains/safety/rigor.ts`, `src/domains/safety/finish-contract.ts` | Advisory validation-contract patterns for scientific artifacts and HPC assumptions. |
 | [evolution.md](evolution.md) | `src/domains/evolution/**`, `src/cli/evolve.ts` | Falsifiable Change Manifest JSON templates, evidence-linked validation, and `clio-coder evolve`. |
+| [gui/README.md](../gui/README.md) | `apps/clio-coder-gui/**`, `dist/gui/**` | Early GUI preview reference: browser documentation, workspace inspection, REST routes, and parity roadmap. |
+| [boundary-doctrine.md](../gui/boundary-doctrine.md) | `apps/clio-coder-gui/**` | Architectural invariants and boundary principles governing GUI client-server interactions. |
+| [chat-rendering-spec.md](../gui/chat-rendering-spec.md) | `apps/clio-coder-gui/**` | Conversation view blueprint, card taxonomy, diff rendering, and inspector presentations. |
+| [cli-surface-routing.md](../gui/cli-surface-routing.md) | `apps/clio-coder-gui/**`, `src/cli/**` | Mapping of CLI commands to GUI routing destinations and REST adapters. |
+| [delivery-order.md](../gui/delivery-order.md) | `apps/clio-coder-gui/**` | Ordered delivery sequence and implementation dependencies for GUI capabilities. |
+| [event-bus-coverage.md](../gui/event-bus-coverage.md) | `apps/clio-coder-gui/**`, `src/core/bus-events.ts` | Event taxonomy and ACP boundary filtering for real-time frontend updates. |
+| [inspector-presentation.md](../gui/inspector-presentation.md) | `apps/clio-coder-gui/**` | Inspector layouts, card ordering, and presentation models for telemetry and state. |
+| [parity/01-cli-surface.md](../gui/parity/01-cli-surface.md) | `apps/clio-coder-gui/**`, `src/cli/**` | Command-line surface parity audit and route coverage status. |
+| [parity/02-slash-and-surfaces.md](../gui/parity/02-slash-and-surfaces.md) | `apps/clio-coder-gui/**`, `src/interactive/slash-commands.ts` | Interactive slash command and overlay parity audit. |
+| [parity/03-domains-and-plan.md](../gui/parity/03-domains-and-plan.md) | `apps/clio-coder-gui/**` | Runtime domain roster and GUI implementation roadmap. |
+| [performance-baseline.md](../gui/performance-baseline.md) | `apps/clio-coder-gui/**` | Rendering benchmarks, latency baselines, and performance budgets. |
+| [settings-write-surface.md](../gui/settings-write-surface.md) | `apps/clio-coder-gui/**`, `src/core/settings-controls.ts` | Schema v2 settings write coverage and form generation. |
+| [workflow-coverage.md](../gui/workflow-coverage.md) | `apps/clio-coder-gui/**` | Operator workflow matrix and human interaction patterns. |
+| [battletest-2026-09-20-fixes.md](battletest-2026-09-20-fixes.md) | `src/**`, `tests/**` | Historical record of the Pi 0.86.1 battletest verification and defect resolutions. |
+| [configure-tree-proposal.md](configure-tree-proposal.md) | `src/core/defaults.ts` | Historical design proposal for configure and `/settings` information architecture. |
 | [config-knobs-audit.md](../history/config-knobs-audit.md) | `src/cli/config.ts` | Point-in-time inventory of legacy environment variables (Historical Appendix). |
 
 ---

package/docs/process/eval-runner.md

@@ -243,7 +243,7 @@ These observations also do not reconcile the separate `trackedMetrics` ledger se
 
 ---
 
-Full reconciliation across session, stdout, fork and out-of-turn evidence is deferred to v0.4.5 or later. Version 0.4.3 does not add an automatic rejection of cost or efficiency comparisons merely because those sources are partial or mixed. Matching source counts do not prove complete coverage or shared call identity. Existing missing-metric, serving-configuration and execution-envelope comparison gates still apply.
+In v0.5.0, full reconciliation across session, stdout, fork, and out-of-turn evidence remains an architectural limitation; eval tracked/stdout folds omit the failed-compaction sidecar. Evaluation runs do not automatically reject cost or efficiency comparisons merely because those sources are partial or mixed. Matching source counts do not prove complete coverage or shared call identity. Existing missing-metric, serving-configuration, and execution-envelope comparison gates still apply.
 
 ## Eval Artifact Format (v4)
 
@@ -314,8 +314,7 @@ mutable external dataset:
   dispatch through the production gate, runs a scripted worker, and verifies
   the sealed receipt and result-contract outcome. The 26 scenarios require no
   model; they do not infer behavior by grepping recipe frontmatter.
-- `behavioral-model.yaml` provides four isolated main-agent scenarios on the
-  `mini` target: a focused edit, adversarial scope control, required
+- `behavioral-model.yaml` provides nine isolated model scenarios, including: a focused edit, adversarial scope control, required
   delegation, and recovery after Bash is denied. Together they cover all eight
   behavioral categories with per-tool call and blocked-call counts, distinct
   and allowlisted read-path counts, declared decoy hits, and grader-emitted
@@ -326,9 +325,9 @@ mutable external dataset:
   that the rules can reject observed model behavior rather than merely restate
   aggregate success counters.
 
-These are source-checkout workflows: the npm archive keeps the inputs for
-inspection and reproducibility, but the deterministic TypeScript driver uses
-the repository development toolchain. Build once, then run either focused
+These are source-checkout workflows and are excluded from the npm archive.
+The deterministic driver uses the shared test harness and repository development
+toolchain; the production evaluation engine remains part of the installed CLI. Build once, then run either focused
 suite from the repository root:
 
 ```sh
@@ -464,15 +463,15 @@ metric means and variances. When the prompt or recipe identity changes, the
 generated evidence names each affected corpus scenario and role instead of
 hiding it behind an aggregate score.
 
-### Reference behavioral baseline
+### Behavioral results
 
-`evals/behavioral-machinery-baseline.json` is retained as reviewable reference
-evidence for the machinery corpus. It is not a CI or release gate. Run the
-current `evals/behavioral-machinery.yaml` through the built CLI when a prompt,
-recipe, policy, or expected-behavior change needs a fresh measurement, inspect
-the named scenario evidence, and update any retained baseline deliberately in
-the reviewed change. Model-required and negative-control suites remain manual
-measurements tied to their exact target and serving configuration.
+Generate current measurements with `evals/behavioral-machinery.yaml` when a
+prompt, recipe, policy or expected behavior changes. Inspect named scenario
+evidence and retain results outside the source tree. The former checked-in
+baseline was historical output rather than a release gate and remains in Git
+history. Live model and negative-control suites are explicit measurements tied
+to the selected target and serving configuration. See `evals/README.md` for
+suite prerequisites, including the fixed-source tracked-metrics campaign.
 
 ### Hard thresholds and informational budgets
 
@@ -523,6 +522,51 @@ The current reference `mini` endpoint is the llama.cpp router at `192.168.86.141
 
 Task outcome commands declared under `verify.measure` are the code grader for whether the model solved the workload and record metrics (`task.solved`, `task.exitCode`). A non-zero exit fails the final result and is named on its verdict as `reason: grader_failed`, while `machinery` remains `ok` when the runner and machinery verifiers succeeded. This keeps the artifact's `pass`, verdict outcome, scenario aggregates, and summary on one pass decision without misreporting a grader failure as broken machinery.
 
+### Grader metric lines (`clio-coder.eval.measure.v1`)
+
+A measure command may also print JSON lines on stdout carrying `"schema": "clio-coder.eval.measure.v1"` and a `metrics` object. Every stdout line is parsed, including lines printed before a nonzero exit. A key is kept only when it is one of these and its value has the listed shape:
+
+- `claims.unsupported` or `completion.reported`, the two behavior keys, with a finite number or a boolean;
+- a key that starts with `custom.digest.`, has at least one character after that prefix, and matches `^[A-Za-z0-9._-]{1,128}$`, with a lowercase hex SHA-256 string matching `^[0-9a-f]{64}$`;
+- any other key that starts with `custom.`, has at least one character after that prefix, and matches `^[A-Za-z0-9._-]{1,128}$`, with a finite number or a boolean.
+
+Every other key or value is dropped, including a number under `custom.digest.` and a string anywhere else. The digest form is how a behavior digest reaches the sealed artifact. Grader metrics are merged into the result last, so the `custom.` prefix is what stops a grader from overwriting `task.solved`, `patch.*`, `tokens.*`, `receipt.*`, or `tools.*`. A `custom.` key whose name the artifact redactor matches, such as `custom.io_token_bytes`, fails the item as `command_error` with an error naming the key. Storage would otherwise replace its value with the string `[redacted]`.
+
+```json
+{"schema":"clio-coder.eval.measure.v1","metrics":{"custom.io.bytes_read":1234,"custom.cache_hit":true,"custom.digest.behavior":"0123456789abcdef0123456789abcdef0123456789abcdef0123456789abcdef"}}
+```
+
+That line lands in the trial's own `results[].metrics` with the three values unchanged. Within one trial the last measure line wins for each key, so a suite that needs one value per scenario uses one eval task per scenario. `verify.assertions` and suite `thresholds` resolve a custom key by exact name, and an `eq` or `neq` assertion compares a digest against a string literal. An assertion on a key the grader did not print fails closed as `assertion_unresolved`. Aggregates and `eval compare` do not read custom keys; a consumer computes its own statistics from the per-trial values.
+
+The runner keeps at most 200,000 characters of each measure command's stdout: the first 20,000 and roughly the last 180,000. A metric line in the dropped middle is lost without an error, and a line cut at either boundary does not parse. Print metric lines last, or keep grader stdout under the limit.
+
 ## Artifact reporting and comparison
 
 Eval runs print a report command for custom output artifacts. `clio-coder eval report --artifact <path>` reads the original artifact with strict provenance validation and does not import it into the store. Comparison aligns jointly declared target/model variations while preserving both route identities, rejects ambiguous route groups, and retains execution-envelope differences. Compact mismatch and hard-gate identities omit runner attachments; full artifact reports retain them.
+
+
+### Interrupted suites and negative probes
+
+Each completed matrix item is atomically checkpointed under the data directory at
+`evals/<eval-id>/checkpoint.json`, before the next item starts. The checkpoint
+records the suite identity, planned and completed counts, status, and graded
+results. A `running` checkpoint is partial evidence, not a completed suite report;
+unstarted cases have no result. The normal artifact is still written at completion.
+
+Graders read a complete, private runner event file rather than the bounded console
+excerpt. Capture fails closed above 64 MiB; a truncated stream cannot silently
+produce a passing grade. This file is retained alongside the session ledgers.
+
+Measure commands retain `measureStdout` and `measureStderr` in result artifacts,
+subject to normal artifact redaction. The shipped `proposal.*`, `memory.*`, and
+`scope.*`, and `explanation.*` grader facts are admitted alongside `custom.*`, `claims.unsupported`, and
+`completion.reported`. Reserved host metrics cannot be overwritten by a grader.
+Gateway call metrics retain both the outer gateway call and the named capability's
+outcome; aggregate tool calls still count one model invocation. Gateway discovery
+and description are not capability executions.
+
+An explicitly negative task, such as checking that an unsupported URL is rejected,
+may set `verify.allowNoop: true`. This bypasses only the no-op gate, and only when
+an independent measure command records `task.solved: true`. Runner errors, blocked
+headless completion, grader failures, and other verification failures still fail.
+Ordinary tasks retain the default no-op failure behavior.

package/docs/process/performance-methodology.md

@@ -446,9 +446,17 @@ latency.
 
 The instant-shell slice uses one `TerminalLease`; Stage 1 does not start a
 second terminal or reconstruct editor state. The built Stage 0 static closure
-is 5 JavaScript chunks and 90,654 bytes. Its regression limit is 6 chunks and
-110,000 bytes, and the closure must contain no orchestrator, provider, tool,
-codewiki, tree-sitter, or Pi implementation marker. The pre-Stage 0 target/auth
+is measured from the production esbuild metafile by
+`tests/contracts/instant-shell-import-graph.test.ts`, after `pnpm build`.
+With bundled Pi TUI 0.86.1, the measured baseline is 13 JavaScript chunks and
+666,307 bytes, including 152,325 bytes attributed to Clio source. The regression
+limits are 16 chunks, 700,000 total bytes, and 165,000 Clio source bytes.
+The separate Clio cap prevents the vendor bundle from masking application growth.
+The closure excludes the orchestrator, provider/agent engines, tools, workers,
+and tree-sitter; Pi's TUI implementation is intentionally bundled and counted.
+The former 5-chunk/90,654-byte measurement and 6-chunk/110,000-byte limit described
+an older build without bundled Pi TUI and had no corresponding artifact test.
+These are bundle-size gates, not a new wall-clock startup benchmark. The pre-Stage 0 target/auth
 check uses a data-only runtime manifest, checked against every canonical
 built-in descriptor, and a read-only credential-presence path; it evaluates no
 provider implementation or OAuth flow.

package/docs/process/release-cut-checklist.md

@@ -16,6 +16,31 @@ implicitly repeated by a publish lifecycle hook.
 4. Install once with `pnpm install --frozen-lockfile`. Record cold installation time
    separately when comparing hosted gate times.
 
+## Working-tree pre-release checks
+
+Before the final candidate is committed, use the existing development commands:
+
+```bash
+pnpm run ci
+node scripts/check-release.mjs
+pnpm run test:package
+```
+
+Run relevant extended provider, configuration, and terminal checks when those
+surfaces changed. These working-tree results prepare a candidate; they do not
+create the publication qualification receipt. Do not invoke `ci:release` on a
+dirty tree: it requires a clean commit and invalidates the previous receipt before
+checking the source. Qualify again after the remaining implementation work is
+finished and the final candidate is committed.
+
+For v0.5.0, inspect configure and `/settings` navigation, effective-settings
+privacy, footer diagnostics, passive target discovery, explicit reasoning/tool
+qualification, and direct/gateway cancellation and lifecycle behavior. Treat live
+provider smokes as resource-consuming checks and coordinate server access. A
+metadata listing does not prove tool correctness, full-context capacity, or GPU
+residency. Keep platform coverage and unresolved process-isolation limitations
+explicit; a successful smoke is not evidence of OS containment.
+
 ## Qualify the candidate
 
 ```bash
@@ -35,7 +60,8 @@ This runs:
   eval first-call selection.
 
 No real model, npm login or publication occurs. An unavailable registry or advisory
-service fails qualification. Chrome must be installed at `/usr/bin/google-chrome` on
+service fails qualification. Native search contracts require `rg` and `fd` (or
+`fdfind`) on `PATH`; the hosted qualification job installs both. Chrome must be installed at `/usr/bin/google-chrome` on
 Linux. Windows subprocess semantics have a separate narrow CI lane; a Linux pass is
 not evidence of a full Windows or macOS pass.
 
@@ -104,8 +130,8 @@ work is proven on canonical main. Published tags are immutable.
 ```bash
 pnpm run test:file -- tests/extended/<file>.test.ts
 pnpm run test:full
-pnpm run test:web:full
-pnpm --filter @iowarp/clio-coder-web verify
+pnpm run test:gui:full
+pnpm --filter @iowarp/clio-coder-gui verify
 pnpm run smoke:real-home -- --target <configured-target-id> --strict
 ```
 

package/docs/process/tool-audit-v0.4.9.md

@@ -1,5 +1,8 @@
 # Individual Tool Audit for v0.4.9
 
+> [!NOTE]
+> **Historical Record**: This document is a point-in-time architectural audit of tool placement, contracts, and dispositions conducted during the v0.4.9 cycle. For active, authoritative tool contracts and argument specifications, refer to [Tool Usage](../guide/tool-usage.md) and [Prompt Envelope and Tools](../architecture/prompt-envelope-and-tools.md).
+
 This is the landed disposition of each builtin tool and each dynamic capability family. It replaces the pre-implementation findings with the current contract, without treating placement as proof of correctness. Source references name the implementation; [Tool usage](../guide/tool-usage.md) supplies detailed arguments and examples. `src/tools/surface.ts` owns placement and `src/tools/policy.ts` owns action class/concurrency. There are 30 builtin names, plus extension and MCP families whose individual schemas are discovered at runtime.
 
 The implementation commits are on the candidate branch; this document does not claim that the version has been published. Scientific integrity below describes evidence and limits, not an empirical certification of each tool. Existing orchestration and interaction behavior was checked for its documentation contract; no new scientific benchmark or full source re-audit is claimed.

package/examples/extensions/README.md

@@ -2,9 +2,7 @@
 
 These installable examples use the harness extension lifecycle, separately from
 library recipes. They require a Clio build containing operator runtime API v1.
-The manifests retain the current package compatibility minimum, 0.4.7; use this
-implementation or a later release that includes it (the original 0.4.7 release
-predates the runtime field).
+The examples require Clio Coder 0.5.0 or later; their manifests enforce that minimum.
 
 From the repository root:
 

package/examples/extensions/lab-status/clio-coder-extension.json

@@ -3,7 +3,7 @@
 	"name": "Local experiment status",
 	"version": "1.0.0",
 	"description": "Inspect a local experiment record; default data is explicitly synthetic.",
-	"compatibility": { "clio": ">=0.4.7" },
+	"compatibility": { "clio": ">=0.5.0" },
 	"runtime": {
 		"api": 1,
 		"entrypoint": "extension.mjs",

package/examples/extensions/measurements/clio-coder-extension.json

@@ -3,7 +3,7 @@
 	"name": "Measurement summary",
 	"version": "1.0.0",
 	"description": "Count, mean, min, max and sample standard deviation of supplied finite measurements.",
-	"compatibility": { "clio": ">=0.4.7" },
+	"compatibility": { "clio": ">=0.5.0" },
 	"capabilities": {
 		"tools": [
 			{

package/library/registry.yaml

@@ -187,7 +187,7 @@ entries:
       - Clio mock provider harness
     version: 0.3.1
     sourceUrl: skills/meta/clio-coder-test
-    sha256: fc8b0f3f7748070456bcf5366a0cf0bb41bf1737549592d3064b7b81cc9b0512
+    sha256: c5a1fa81a19d671560d98a5f4e78f56a0515294611bc9ae796aac37165719a92
     provides:
       - kind: skill
         name: clio-coder-test
@@ -426,9 +426,9 @@ entries:
       - resolve the bug in issue
       - implement this tracker issue
       - fix a GitHub issue end to end
-    version: 0.3.0
+    version: 0.3.1
     sourceUrl: skills/git/fix-issue
-    sha256: 1a2bae6bebcde3951e09d32d6abfc3bf85132da915b8e1c4295850fe9fc94b57
+    sha256: 2ac7dd53397fff8a12f829f3e9241c839c08e23dcf80d435c323add6b0cd03c2
     provides:
       - kind: skill
         name: fix-issue
@@ -770,9 +770,9 @@ entries:
       - commit and open the PR
       - push and open a pull request
       - get this up for review
-    version: 0.5.0
+    version: 0.5.1
     sourceUrl: skills/git/ship
-    sha256: a70df2c1f3546262557372ece5733722f6eb737b85ec92902fcc0b5cf94cc123
+    sha256: cb65a5b59d98ac2d6105a9aa52446c7e4a97c33a9a5a240d74f82f2447b56534
     provides:
       - kind: skill
         name: ship
@@ -802,6 +802,31 @@ entries:
         description: Writes, reviews, or prunes a SKILL.md, judging whether its
           description, body, and length earn their cost. Not for packaging a
           workflow that just happened; use workflow-distiller.
+  - kind: skill
+    name: slurm-jobs
+    description: Submits a batch script to Slurm through the clio-kit Slurm MCP
+      server, polls it to a terminal state, fetches its output, and stamps the
+      scheduler job id into the final answer. Not for writing the science inside
+      the script or for interactive allocations.
+    category: research
+    audit: pass
+    triggers:
+      - submit this job to slurm
+      - run this on the cluster
+      - sbatch this script
+      - check my slurm job
+      - cancel my slurm job
+      - is the queue busy
+    version: 0.1.0
+    sourceUrl: skills/research/slurm-jobs
+    sha256: d57b83965e1a1dbc46adbad605d755a7a69368c4f4e63e1e181a416a876dce40
+    provides:
+      - kind: skill
+        name: slurm-jobs
+        description: Submits a batch script to Slurm through the clio-kit Slurm MCP
+          server, polls it to a terminal state, fetches its output, and stamps
+          the scheduler job id into the final answer. Not for writing the
+          science inside the script or for interactive allocations.
   - kind: skill
     name: tdd
     description: Builds a feature or fix test-first, red to green at pre-agreed

package/library/skills/git/fix-issue/SKILL.md

@@ -7,7 +7,7 @@ triggers:
   - resolve the bug in issue
   - implement this tracker issue
   - fix a GitHub issue end to end
-version: 0.3.0
+version: 0.3.1
 license: Apache-2.0
 compatibility: git >=2.30.0, gh CLI >=2.0.0 (authenticated for issue viewing), POSIX-compatible shell
 allowed-tools:
@@ -17,6 +17,7 @@ allowed-tools:
   - ls
   - git
   - bash
+  - verify
   - context
   - code_nav
   - dispatch

package/library/skills/git/fix-issue/plugin.json

@@ -1,7 +1,7 @@
 {
 	"$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
 	"name": "fix-issue",
-	"version": "0.3.0",
+	"version": "0.3.1",
 	"description": "Resolves a tracker issue end to end: fetches it, diagnoses only when the issue does not name the code, fixes test-first, and self-reviews against the acceptance criteria, leaving a verified uncommitted change. Not for filing; use file-ticket. Not for committing or PRs; use ship.",
 	"license": "Apache-2.0",
 	"extensions": {

package/library/skills/git/ship/SKILL.md

@@ -7,7 +7,7 @@ triggers:
   - commit and open the PR
   - push and open a pull request
   - get this up for review
-version: 0.5.0
+version: 0.5.1
 license: Apache-2.0
 compatibility: git >=2.30.0, gh CLI >=2.0.0 (required for pr mode), POSIX-compatible shell
 allowed-tools:
@@ -16,6 +16,7 @@ allowed-tools:
   - ls
   - git
   - bash
+  - verify
   - tasks
   - ask_user
 clio-coder:

package/library/skills/git/ship/plugin.json

@@ -1,7 +1,7 @@
 {
 	"$schema": "https://agent-plugins.org/schemas/1.0.0/plugin.schema.json",
 	"name": "ship",
-	"version": "0.5.0",
+	"version": "0.5.1",
 	"description": "Ships finished work: writes one reviewed atomic commit, keeps maintainer branches local, or pushes a contributor branch to their fork and opens a PR only on explicit intent. Not for producing the change; use fix-issue.",
 	"license": "Apache-2.0",
 	"extensions": {

package/library/skills/meta/clio-coder-test/SKILL.md

@@ -41,8 +41,8 @@ npm run build                             # tsup plus the codewiki asset
 npm run test:file -- tests/smoke/<file>.test.ts  # requires a current build
 npm run test                              # required contracts and three process smoke files
 npm run test:full                         # explicit extended root investigation
-npm run test:web:full                     # explicit extended web investigation
-pnpm --filter @iowarp/clio-coder-web verify # app types, lint, tests, build, browser
+npm run test:gui:full                     # explicit extended web investigation
+pnpm --filter @iowarp/clio-coder-gui verify # app types, lint, tests, build, browser
 npm run ci                                # deterministic root gate
 npm run ci:release                        # qualify clean committed source and exact installed tarball
 npm run release:preflight                 # fast check of the unchanged qualified package
@@ -54,9 +54,8 @@ boot/reconnect against the installed server. The full web `verify` command is an
 explicit development investigation, including its viewport/accessibility matrix.
 Do not run full development tests repeatedly during release or publication.
 
-The separate trace viewer is retired. `apps/workbench/` is retained reference
-source, excluded from builds, publication and product gates. Use `test:web` for
-the required app tests alone; app `verify` requires headless Chrome.
+The separate trace viewer and the earlier workbench client are both retired. Use
+`test:gui` for the required app tests alone; app `verify` requires headless Chrome.
 
 No deterministic gate contacts a real model. When a task explicitly requires
 live validation, build first and run `node dist/cli/index.js run` against a
@@ -73,7 +72,7 @@ settings with the result.
 | Any import edit under `src/` | `npm run lint` | Hygiene invokes all six boundary rules. |
 | CLI, entry, process lifecycle, or ACP stdio flow | Build, then the closest file under `tests/smoke/` | Smoke executes the built binary. |
 | Published package contents | Build, then `tests/smoke/installed-package.test.ts` | The test packs and installs the actual artifact. |
-| Unified web application | `npm run test:web` | Required authentication/runtime boundaries; full `verify` is optional development work. |
+| Unified web application | `npm run test:gui` | Required authentication/runtime boundaries; full `verify` is optional development work. |
 
 Read `references/test-map.md` for the current file map and exact subset
 commands.

package/library/skills/meta/clio-coder-test/evals.md

@@ -51,10 +51,10 @@ One representative scenario via `clio-coder eval skill` against Nemo-3.5-Lightni
 Prompt: "The web API changed. Which checks belong to this release?"
 Expected:
 - Builds the root CLI for ACP fixtures, then runs
-  `pnpm --filter @iowarp/clio-coder-web verify` with temporary state outside the checkout.
+  `pnpm --filter @iowarp/clio-coder-gui verify` with temporary state outside the checkout.
 - Names app type, API/worker boundary, test, client build and headless-browser checks.
 - Explains root `ci:release` includes web tests and the installed package audit.
-- Does not run the retired trace viewer or reference-only Workbench gates.
+- Does not run the retired trace viewer or workbench gates.
 
 RED: the prior 0.3.0 instructions prescribe both obsolete application lanes.
 GREEN: the current instructions name only the unified app lane and retain the

package/library/skills/meta/clio-coder-test/references/test-map.md

@@ -13,8 +13,8 @@ Extended regressions are explicit development investigations.
 | Package qualification | Installed package and native call timing under `tests/smoke/` | `npm run ci:release` | Built once by qualification |
 | Extended regressions | `tests/extended/`, `tests/extended-smoke/` | `npm run test:full` or a focused `test:file` | Yes for CLI scenarios |
 | Static boundaries and pins | `scripts/check-hygiene.ts` | `npm run lint` | No |
-| Required web boundaries | Selected files in `apps/clio-coder-web/tests/` | `npm run test:web` | Root build |
-| Full web investigation | All web tests and browser matrix | `pnpm --filter @iowarp/clio-coder-web verify` | Yes |
+| Required GUI boundaries | Selected files in `apps/clio-coder-gui/tests/` | `npm run test:gui` | Root build |
+| Full GUI investigation | All GUI tests and browser matrix | `pnpm --filter @iowarp/clio-coder-gui verify` | Yes |
 
 `npm run ci` runs the routine lane. `npm run ci:release` qualifies a clean,
 committed candidate and its exact tarball; `npm run release:preflight` checks

package/library/skills/registry.yaml

@@ -40,16 +40,16 @@ skills:
     sha256: 65029c2d9d545728750c8a713f92035bd6bdbfa9874de442df72bd059846cb7a
   - name: fix-issue
     path: git/fix-issue
-    version: 0.3.0
-    sha256: 626e3aa8ca95bc603f2e0cdd2e7502aa96be85af6bfa92b635c39ef99d3845a5
+    version: 0.3.1
+    sha256: 0e3aa19b6214e03bf581d29b097f99a3e1110d4303a1598c8a996171b8606dc8
   - name: resolve-merge-conflicts
     path: git/resolve-merge-conflicts
     version: 0.4.0
     sha256: 901f482aa2231552d64a721e0d1c58dc8f211347543b5a91fd2ab3c7e413c11c
   - name: ship
     path: git/ship
-    version: 0.5.0
-    sha256: 4d626718fdb8b67cf3f22d09ac5a228e5c8103befa1295a672587a6605f038b1
+    version: 0.5.1
+    sha256: 07bc1acdf3506c7d8a720abe67144a32bb8e18132e080a3b9aaf66cba267876c
   - name: worktree-create
     path: git/worktree-create
     version: 0.6.0
@@ -66,7 +66,7 @@ skills:
   - name: clio-coder-test
     path: meta/clio-coder-test
     version: 0.3.1
-    sha256: 1e8c839b3668e50cd8ed7c69e7ae3ccc5afa15a1db72eb2b8babb1672134fba1
+    sha256: e5de4cd00f6ca78aa6443b13d54751ef1088ece0294d4f24f418e037dabf2214
   - name: credentials
     path: meta/credentials
     version: 0.2.0
@@ -125,6 +125,10 @@ skills:
     path: research/scientific-modernization
     version: 0.4.1
     sha256: eb31099d764cba913c19726f6ba85d7081c5e01c81a3b2cd45eb136a4e47b858
+  - name: slurm-jobs
+    path: research/slurm-jobs
+    version: 0.1.0
+    sha256: 7692e523512f56e65f881ff87c2e997b8baf4d487e5b29f831e62dc97dc0e1c3
   # ── workflow ──
   - name: cut-it
     path: workflow/cut-it

package/library/skills/research/slurm-jobs/SKILL.md

@@ -0,0 +1,110 @@
+---
+name: slurm-jobs
+description: "Submits a batch script to Slurm through the clio-kit Slurm MCP server, polls it to a terminal state, fetches its output, and stamps the scheduler job id into the final answer. Not for writing the science inside the script or for interactive allocations."
+triggers:
+  - submit this job to slurm
+  - run this on the cluster
+  - sbatch this script
+  - check my slurm job
+  - cancel my slurm job
+  - is the queue busy
+version: 0.1.0
+license: Apache-2.0
+clio-coder:
+  registry-id: iowarp/clio-coder
+  source-url: https://github.com/iowarp/clio-coder/tree/main/library/skills/research/slurm-jobs
+  audit: pass
+  provenance: designed
+  eval-status: smoke-checked
+  model-size: any
+---
+
+# Slurm Jobs
+
+Run one batch job end to end through the five `mcp_slurm__*` gateway
+capabilities, and leave a record another person can check: the job id, the
+terminal state, and where the output is. A job that was submitted but never
+described to a terminal state is not a result.
+
+Anti-trigger: if the question is what the script should compute, or why its
+numbers are wrong, settle that first; use experiment-protocol or
+scientific-debugging. This skill moves a finished script through the
+scheduler.
+
+## Arguments
+
+```text
+/skill slurm-jobs <script path, resources, and what to report>
+```
+
+## Phase 0 - Prerequisite check
+
+1. `gateway(op="find", query="slurm")`. The listing must show a server with id
+   `slurm` and the capabilities `mcp_slurm__slurm_submit`, `mcp_slurm__slurm_list`,
+   `mcp_slurm__slurm_describe`, `mcp_slurm__slurm_cluster`, and
+   `mcp_slurm__slurm_cancel`.
+2. If the server is missing, stop and tell the operator to declare it (the
+   entry is in the Slurm guide, `docs/guide/slurm.md`) and to run
+   `clio-coder doctor`, which reports `clio-kit`, the declaration, and
+   `sbatch`. If it is listed `untrusted` or `stale`, give the remedy the
+   listing prints. Do not fall back to `bash` with `sbatch`: the operator chose
+   the mediated path.
+3. `gateway(op="describe", capability="mcp_slurm__slurm_submit")` once, and
+   use the parameter names it returns. Do not guess them.
+
+Every call to these capabilities asks the operator for approval, reads
+included, because an MCP server carries one action class for all of its
+tools. Poll sparingly. In a headless run the approval is denied: report that
+the run needs an operator and stop.
+
+## Phase 1 - Look before submitting
+
+`mcp_slurm__slurm_cluster` with the target partition, no node details. Confirm
+the partition exists and is up. Read the script with `read` and check that its
+`#SBATCH` lines agree with the resources you are about to request; the call's
+`cores`, `memory`, `time_limit`, and `partition` win over the script, so a
+disagreement is a bug to fix now.
+
+## Phase 2 - Submit
+
+`mcp_slurm__slurm_submit` with `script_path` and explicit `cores`, `memory`,
+`time_limit`, `job_name`, and `partition`. Set `array` only for an array job.
+Record the returned job id verbatim at once, in your working notes and in the
+`tasks` board if one is open. Submission is not idempotent: never resubmit
+because a reply was slow. Look for the job with `mcp_slurm__slurm_list`
+filtered by your user and the job name first.
+
+## Phase 3 - Poll to a terminal state
+
+`mcp_slurm__slurm_describe` with the job id and `output="none"`. The result
+says whether the state is terminal. While it is not, wait before asking again:
+start at 30 seconds and double up to 5 minutes. Do not spin. A job pending
+for longer than the operator's patience is a finding to report with the
+pending reason the scheduler gives, not something to cancel on your own.
+
+## Phase 4 - Fetch the output
+
+Once terminal, `mcp_slurm__slurm_describe` with `output="both"` and a
+`max_output_chars` that fits the turn. The tails are bounded; for full output
+read the files the description names with `read`. A `COMPLETED` state with a
+nonzero exit code, or an empty stdout where output was expected, is a failure.
+Say so.
+
+## Phase 5 - Cancel only when asked
+
+`mcp_slurm__slurm_cancel` takes `job_id` and `confirm_job_id`, which must
+repeat it exactly. Cancel only a job the operator named or one you submitted
+in this session and were told to stop. Describe the job afterwards to confirm
+it reached `CANCELLED`.
+
+## Final answer
+
+End with a line in exactly this form, one per job:
+
+```text
+Slurm job <job id>: <terminal state>, exit <code>, stdout <path>, stderr <path>
+```
+
+If the job never reached a terminal state, write `Slurm job <job id>: <state>
+(not terminal)` and say what the operator should check next. Never report
+success without the job id.