dsh-harbor-evolution 0.7.1 → 0.7.2

package/README.md

@@ -14,10 +14,10 @@ npx --yes dsh-harbor-evolution@latest setup --project-root "$PWD"
 
 The setup command installs both required runtimes:
 
-- `harbor-dsh-evolution==0.7.1` in a managed Python environment.
-- `dsh-harbor-evolution@0.7.1` in the selected DSH profile.
+- `harbor-dsh-evolution==0.7.2` in a managed Python environment.
+- `dsh-harbor-evolution@0.7.2` in the selected DSH profile.
 
-It then stores the absolute Harbor executable paths and `projectRoot` in the profile's `harbor-evolution` block and verifies the integration. Existing unrelated profile entries are preserved, and rerunning setup updates the same block.
+It then stores the absolute Harbor executable paths and a fallback `projectRoot` in the profile's `harbor-evolution` block and verifies the integration. Agent Tool calls always use the calling session's absolute working directory as their project root; the configured value remains the Web Workbench and non-Agent fallback. Existing unrelated profile entries are preserved, and rerunning setup updates the same block.
 
 The default profile is `web`. Use `--profile headless` only when that is the profile you actually run. See all options with:
 
@@ -79,7 +79,7 @@ The selected profile receives one id-targeted override:
     pythonPath: ""
 ```
 
-Keep `pythonPath` empty for the published Python package. `candidatePath`, `datasetPath`, `jobPath`, and `policyPath` are constrained to `projectRoot`.
+Keep `pythonPath` empty for the published Python package. For Agent Tool calls, `projectRoot` is replaced by the calling session's working directory for that call. `candidatePath`, `datasetPath`, `jobPath`, and `policyPath` remain constrained to that request-local root, so concurrent sessions cannot redirect each other's Harbor operations.
 
 For source development from the repository:
 

package/index.js

@@ -47,12 +47,20 @@ function jsonTool(definition, execute) {
       schema: { type: 'string' },
       render: (_args, value) => [{ type: 'text', text: value }],
     },
-    async execute(args) {
-      return JSON.stringify(await execute(args), null, 2)
+    async execute(args, exec) {
+      return JSON.stringify(await execute(args, exec), null, 2)
     },
   })
 }
 
+function toolProjectRoot(exec) {
+  const cwd = exec?.agent?.session?.header?.cwd
+  if (typeof cwd !== 'string' || !path.isAbsolute(cwd)) {
+    throw new Error('Harbor tools require an Agent session with an absolute working directory')
+  }
+  return path.resolve(cwd)
+}
+
 export function apply(ctx, config) {
   const resolved = {
     ...config,
@@ -66,7 +74,12 @@ export function apply(ctx, config) {
     ),
   }
   const modelRuntime = new CandidateModelRuntime(ctx, resolved)
-  const service = new EvolutionService(resolved, { pluginVersion: packageJson.version }, modelRuntime)
+  const metadata = { pluginVersion: packageJson.version }
+  const service = new EvolutionService(resolved, metadata, modelRuntime)
+  const serviceForTool = exec => new EvolutionService({
+    ...resolved,
+    projectRoot: toolProjectRoot(exec),
+  }, metadata, modelRuntime)
 
   ctx.skills.register(loadBundledSkill())
   installDashboardWeb(ctx, service)
@@ -79,7 +92,7 @@ export function apply(ctx, config) {
       candidateId: { type: 'string' },
       version: { type: 'string' },
     },
-  }, args => service.snapshot(args)))
+  }, (args, exec) => serviceForTool(exec).snapshot(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_evolution_init',
@@ -101,7 +114,7 @@ export function apply(ctx, config) {
       policyVersion: { type: 'string', required: true },
       minImprovement: { type: 'number', required: true },
     },
-  }, args => service.initialize(args)))
+  }, (args, exec) => serviceForTool(exec).initialize(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_evolution_doctor',
@@ -116,7 +129,7 @@ export function apply(ctx, config) {
       candidateModel: { type: 'string' },
       candidateReasoningEffort: { type: 'string' },
     },
-  }, args => service.doctor(args)))
+  }, (args, exec) => serviceForTool(exec).doctor(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_dataset_validate',
@@ -124,7 +137,7 @@ export function apply(ctx, config) {
     parameters: {
       datasetPath: { type: 'string', required: true },
     },
-  }, args => service.validateDataset(args)))
+  }, (args, exec) => serviceForTool(exec).validateDataset(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_context_preview',
@@ -140,7 +153,7 @@ export function apply(ctx, config) {
       candidateModel: { type: 'string' },
       candidateReasoningEffort: { type: 'string' },
     },
-  }, args => service.previewContext(args)))
+  }, (args, exec) => serviceForTool(exec).previewContext(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_eval_run',
@@ -158,7 +171,7 @@ export function apply(ctx, config) {
       candidateModel: { type: 'string' },
       candidateReasoningEffort: { type: 'string' },
     },
-  }, args => service.run(args)))
+  }, (args, exec) => serviceForTool(exec).run(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_eval_result',
@@ -170,7 +183,7 @@ export function apply(ctx, config) {
       compareJob: { type: 'string', description: 'Optional previous Job for view=governance impact analysis' },
       since: { type: 'string', description: 'Optional ISO timestamp for incremental progress changes' },
     },
-  }, args => service.result(args)))
+  }, (args, exec) => serviceForTool(exec).result(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_evaluator_inspect',
@@ -178,7 +191,7 @@ export function apply(ctx, config) {
     parameters: {
       stackPath: { type: 'string', description: 'Defaults to .harbor/evaluation-stack.yml' },
     },
-  }, args => service.evaluatorInspect(args)))
+  }, (args, exec) => serviceForTool(exec).evaluatorInspect(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_evaluator_update',
@@ -191,7 +204,7 @@ export function apply(ctx, config) {
       newEvaluatorVersion: { type: 'string', required: true },
       newStackVersion: { type: 'string', required: true },
     },
-  }, args => service.evaluator(args)))
+  }, (args, exec) => serviceForTool(exec).evaluator(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_ground_truth_init',
@@ -205,7 +218,7 @@ export function apply(ctx, config) {
       provenance: { type: 'string', required: true },
       criteria: { type: 'string', required: true, description: 'Comma-separated criterion ids' },
     },
-  }, args => service.groundTruthInitialize(args)))
+  }, (args, exec) => serviceForTool(exec).groundTruthInitialize(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_evaluator_meta_evaluate',
@@ -215,7 +228,7 @@ export function apply(ctx, config) {
       observationsPath: { type: 'string', required: true },
       outputPath: { type: 'string', description: 'Defaults to .harbor/meta-evaluation-report.json' },
     },
-  }, args => service.evaluatorMetaEvaluate(args)))
+  }, (args, exec) => serviceForTool(exec).evaluatorMetaEvaluate(args)))
 
   ctx.tools.register(jsonTool({
     name: 'harbor_candidate_compare',
@@ -225,5 +238,5 @@ export function apply(ctx, config) {
       candidateJob: { type: 'string', required: true },
       policyPath: { type: 'string', required: true },
     },
-  }, args => service.compare(args)))
+  }, (args, exec) => serviceForTool(exec).compare(args)))
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-harbor-evolution",
-  "version": "0.7.1",
+  "version": "0.7.2",
   "description": "DeepSeek Harness plugin and bundled Skill for safely evolving Cordis Candidates with Harbor.",
   "type": "module",
   "main": "index.js",

package/skills/evolve-agent-with-harbor/SKILL.md

@@ -33,7 +33,7 @@ Do not turn an inspection or diagnostic request into Agent mutation or deploymen
 
 Inspect the current workspace before asking questions. Look for Agent entry files, package metadata, curl examples, Dataset instructions, existing Harbor configuration/Jobs, tests, and available Codex or Claude Code commands. Reuse reliable findings and say what was inferred; do not ask the user to transcribe information already present in files.
 
-When no Harbor workspace exists, propose `./harbor-evolution/` under the current working directory as the managed evaluation workspace. Keep imported snapshots and generated evaluation files there. If the installed Plugin is configured to a different `projectRoot`, explain the mismatch and propose the exact configuration change before writing anything.
+When no Harbor workspace exists, propose `./harbor-evolution/` under the current session working directory as the managed evaluation workspace. Agent-facing Harbor Tools derive `projectRoot` from the calling session for every invocation and keep imported snapshots and generated evaluation files inside that request-local root. Treat the Plugin's configured `projectRoot` only as the Web Workbench and non-Agent fallback; do not block initialization merely because it differs from the current session working directory.
 
 Ask only for missing parts of the four-concept intake, using the user's language and short examples:
 

package/skills/evolve-agent-with-harbor/evals/evals.json

@@ -33,6 +33,17 @@
         "The response maps the supplied paths and curl into the four-concept confirmation card.",
         "The response does not run initialization or evaluation before confirmation."
       ]
+    },
+    {
+      "id": 4,
+      "prompt": "当前 Session 工作目录是 /Users/mac/Documents/Harness，但 Harbor Plugin 配置里显示的 projectRoot 是 /Users/mac/Library/Application Support/XiaoHui Harness/workspace。请在当前项目下初始化 ./harbor-evolution。",
+      "expected_output": "把当前 Session 工作目录作为 Agent Tool 的项目根目录，继续准备 /Users/mac/Documents/Harness/harbor-evolution 的初始化确认卡；说明配置值只供 Web Workbench 或非 Agent 场景回退使用，不要求修改配置，也不因路径不同而拒绝初始化。",
+      "files": [],
+      "assertions": [
+        "The response treats the calling session working directory as the Agent Tool project root.",
+        "The response does not block initialization or require a projectRoot configuration change because the fallback differs.",
+        "The proposed managed workspace stays under /Users/mac/Documents/Harness."
+      ]
     }
   ]
 }

package/skills/evolve-agent-with-harbor/references/initialization.md

@@ -13,7 +13,7 @@ Keep onboarding anchored on four visible concepts:
 | 评测器（评测标准） (Evaluator) | 怎样算好？ | evaluator curl/path, or natural-language criteria | Evaluator, Rubric, Judge identity, Evaluation Contract |
 | 优化器 (Optimizer) | 谁根据结果改进？ | current Agent, Codex, Claude Code, or local command | Optimizer identity, mutation surface, rollback workflow |
 
-Inspect first and prefill everything reliable. Ask only about missing rows. Use `./harbor-evolution/` as the proposed managed workspace when no project exists, but do not write outside the Plugin's configured `projectRoot`; surface a configuration mismatch before initialization.
+Inspect first and prefill everything reliable. Ask only about missing rows. Use `./harbor-evolution/` as the proposed managed workspace when no project exists. Agent-facing Harbor Tools resolve every path inside the calling session's working directory, so a different configured `projectRoot` is only a Web Workbench or non-Agent fallback and must not block initialization.
 
 Before any write, show the four rows plus inferred workspace, diagnostic/promotion scope, and deferred capabilities. The user may start initialization, modify the card, or open advanced configuration.