@iiwish/agentrecord 0.0.1

package/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 AgentRecord contributors
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

package/README.md

@@ -0,0 +1,109 @@
+# AgentRecord
+
+AgentRecord turns local AI-agent work traces into a user-owned, auditable AI work profile.
+
+It is not a hiring score, leaderboard, certification, or prompt-history viewer. The first product surface is a personal AI work record: what kinds of work you do with agents, how you delegate, how you review, how you verify, and what evidence supports those claims.
+
+## Positioning
+
+AgentRecord is a local-first record layer for AI-native work.
+
+- **For users:** a private, inspectable profile of how they work with Codex, Claude Code, opencode, OpenClaw, Hermes, and other agents.
+- **For agents:** a compact, prompt-ready work profile that future agents can use to understand the user's preferences, strengths, constraints, and collaboration style.
+- **For sharing:** a redacted proof page that shows evidence-backed work identity without exposing raw conversations, prompts, secrets, terminal output, or source code.
+
+## Product Principles
+
+- **User-owned by default:** local files first, export second, hosted services optional.
+- **Evidence over claims:** scores and labels must point back to evidence cards and confidence levels.
+- **No raw trace leakage:** public artifacts never include raw prompts, raw session IDs, secrets, terminal output, or private source files.
+- **Cross-agent model:** individual clients are adapters; the profile model is stable across tools.
+- **Incremental runs:** repeated runs update the profile without rewriting history.
+- **No default hiring decision:** recruiter or hiring views are optional audience layers, not the core product.
+
+## Quickstart
+
+After npm install:
+
+```bash
+npm install -g @iiwish/agentrecord
+agentrecord init --owner <owner>
+agentrecord build --config ./agentrecord.config.json --no-account-usage
+agentrecord validate --config ./agentrecord.config.json
+agentrecord open --config ./agentrecord.config.json
+```
+
+From the source checkout:
+
+```bash
+node src/cli.mjs init --owner <owner>
+node src/cli.mjs build --config ./agentrecord.config.json --no-account-usage
+node src/cli.mjs validate --config ./agentrecord.config.json
+node src/cli.mjs open --config ./agentrecord.config.json
+```
+
+The primary product artifact is `profiles/<owner>/index.html`: a single-file, static AI work passport for local review or redacted sharing. `profile.json` and `evidence.jsonl` are the structured truth sources behind the page, and `profile.md` is a secondary audit draft.
+
+Agent context is opt-in:
+
+```bash
+node src/cli.mjs build --config ./agentrecord.config.json --agent-context
+```
+
+Default builds generate only the self/share artifacts and do not generate recruiter, job-agent, or employment-decision views.
+
+A privacy-safe starter example is available in `examples/basic/`.
+
+## Commands
+
+```bash
+node src/cli.mjs init --dry-run
+node src/cli.mjs init --owner <owner>
+node src/cli.mjs scan --config ./agentrecord.config.json
+node src/cli.mjs build --config ./agentrecord.config.json
+node src/cli.mjs validate --config ./agentrecord.config.json
+node src/cli.mjs open --config ./agentrecord.config.json
+```
+
+Useful overrides:
+
+```bash
+node src/cli.mjs --help
+node src/cli.mjs doctor
+node src/cli.mjs scan --sessions-dir ~/.codex/sessions
+node src/cli.mjs build --owner <owner> --locale zh-CN
+node src/cli.mjs build --config ./agentrecord.config.json --agent-context
+```
+
+## Expected Outputs
+
+```text
+profiles/<owner>/
+  profile.json          Machine-readable AI work profile
+  evidence.jsonl        Redacted evidence cards
+  index.html            Human-readable local/share report
+  profile.md            Markdown report
+  redaction-report.md   Privacy boundary audit
+  run-report.md         Incremental run summary
+  .agentrecord/         Private state, cursors, snapshots
+```
+
+Optional explicit outputs:
+
+```text
+profiles/<owner>/
+  agent-context.md      Public-safe prompt context for future agents
+  agent-context.json    Machine-readable agent context pack
+```
+
+The default config file is `agentrecord.config.json`. Output defaults to `profiles/<owner>/`, and private state defaults to `profiles/<owner>/.agentrecord/`. AgentRecord is local-first by default; it does not upload, publish, or host generated artifacts.
+
+## First Supported Adapter
+
+The first supported adapter should be Codex local sessions, because it gives enough real usage history to validate the core loop quickly. Claude Code, opencode, OpenClaw, and Hermes should follow once the normalized event model is stable.
+
+## Status
+
+AgentRecord is in the v0.1 CLI baseline and uses package version `0.0.1` for the first conservative npm release candidate. The `build` and `validate` commands provide the minimum local loop: generate `profile.json`, `evidence.jsonl`, `index.html`, `profile.md`, `redaction-report.md`, and `run-report.md`, then verify required artifacts, profile shape, evidence references, locale parity, private state, and public-artifact privacy boundaries.
+
+HTML is the first product artifact. `profiles/<owner>/index.html` is the primary human-readable output for local review and sharing, backed by the machine-readable profile and evidence files.

package/examples/basic/README.md

@@ -0,0 +1,23 @@
+# AgentRecord basic example
+
+This example is a privacy-safe starting point for a first local profile run. It does not include raw traces, private sessions, generated `profiles/` output, or real user data.
+
+From the package source:
+
+```bash
+node src/cli.mjs init --owner smoke-owner
+node src/cli.mjs build --config ./agentrecord.config.json --no-account-usage
+node src/cli.mjs validate --config ./agentrecord.config.json
+node src/cli.mjs open --config ./agentrecord.config.json
+```
+
+After installing the npm package:
+
+```bash
+agentrecord init --owner smoke-owner
+agentrecord build --config ./agentrecord.config.json --no-account-usage
+agentrecord validate --config ./agentrecord.config.json
+agentrecord open --config ./agentrecord.config.json
+```
+
+The primary output is `profiles/smoke-owner/index.html`, a single-file static HTML report. Keep raw agent traces and private `.agentrecord/` state local.

package/examples/basic/agentrecord.config.example.json

@@ -0,0 +1,36 @@
+{
+  "schema_version": "agentrecord.config.v0",
+  "owner": "smoke-owner",
+  "owner_display_name": "Smoke Owner",
+  "profiles_dir": "profiles",
+  "output": {
+    "profile_dir": "profiles/smoke-owner"
+  },
+  "codex": {
+    "sessions_dir": "~/.codex/sessions",
+    "session_roots": ["~/.codex/sessions"],
+    "account_usage": {
+      "enabled": false,
+      "timeout_ms": 1500
+    }
+  },
+  "memory": {
+    "enabled": false
+  },
+  "evidence_rules_paths": [
+    "../../references/evidence-rules.json"
+  ],
+  "report": {
+    "locale": "en-US",
+    "fallback_locale": "en-US",
+    "label_mode": "bilingual-compact",
+    "schema_language": "en-US",
+    "audiences": ["self", "share"],
+    "default_audience": "self"
+  },
+  "privacy": {
+    "mode": "strict",
+    "public_session_ids": false,
+    "public_project_paths": false
+  }
+}

package/locales/en-US.json

@@ -0,0 +1,45 @@
+{
+  "schema_version": "agentrecord.locale.v0",
+  "locale": "en-US",
+  "html_lang": "en",
+  "ui": {
+    "title": "AgentRecord Profile",
+    "subtitle": "Local, evidence-backed AI work record",
+    "generated": "Generated",
+    "trace_window": "Trace Window",
+    "sessions": "Sessions",
+    "token_activity": "Token activity",
+    "evidence": "Evidence",
+    "identity": "Work Identity",
+    "role_signals": "Role Signals",
+    "ability_model": "Ability Model",
+    "agent_ledger": "Agent Ledger",
+    "calibration": "Calibration",
+    "privacy_boundary": "Privacy Boundary",
+    "run_metadata": "Run Metadata",
+    "public_safe": "Public-safe summary only. Raw prompts, raw responses, raw session IDs, terminal output, source code, and secrets are excluded.",
+    "not_hiring": "This record does not make employment decisions, external comparison claims, or unsupported seniority claims.",
+    "no_memory": "No curated memory evidence was available; this profile uses activity baseline evidence."
+  },
+  "roles": {
+    "product_builder": "Product Builder",
+    "technical_reviewer": "Technical Reviewer",
+    "agent_operator": "Agent Operator",
+    "shipping_owner": "Shipping Owner",
+    "systems_thinker": "Systems Thinker",
+    "research_synthesizer": "Research Synthesizer",
+    "collaboration_handoff": "Collaboration Handoff"
+  },
+  "abilities": {
+    "goal_framing": "Goal Framing",
+    "context_packaging": "Context Packaging",
+    "agent_delegation": "Agent Delegation",
+    "review_judgment": "Review Judgment",
+    "verification_discipline": "Verification Discipline",
+    "failure_recovery": "Failure Recovery",
+    "scope_control": "Scope Control",
+    "shipping_hygiene": "Shipping Hygiene",
+    "product_judgment": "Product Judgment",
+    "collaboration_handoff": "Collaboration Handoff"
+  }
+}

package/locales/zh-CN.json

@@ -0,0 +1,45 @@
+{
+  "schema_version": "agentrecord.locale.v0",
+  "locale": "zh-CN",
+  "html_lang": "zh-CN",
+  "ui": {
+    "title": "AgentRecord 画像",
+    "subtitle": "本地优先、证据支撑的 AI 工作记录",
+    "generated": "生成时间",
+    "trace_window": "Trace 时间窗",
+    "sessions": "会话数",
+    "token_activity": "Token 活跃度",
+    "evidence": "证据",
+    "identity": "工作身份",
+    "role_signals": "角色信号",
+    "ability_model": "能力模型",
+    "agent_ledger": "Agent 台账",
+    "calibration": "校准说明",
+    "privacy_boundary": "隐私边界",
+    "run_metadata": "运行元数据",
+    "public_safe": "仅包含可公开摘要。原始 prompt、原始回复、原始 session ID、terminal output、source code 和 secrets 均不输出。",
+    "not_hiring": "本记录不做录用决策、外部比较结论或未经校准的资历结论。",
+    "no_memory": "未发现可用的 memory 证据，本画像使用活跃度基线证据。"
+  },
+  "roles": {
+    "product_builder": "产品建造者",
+    "technical_reviewer": "技术评审者",
+    "agent_operator": "Agent 操作者",
+    "shipping_owner": "交付负责人",
+    "systems_thinker": "系统思考者",
+    "research_synthesizer": "研究综合者",
+    "collaboration_handoff": "协作交接者"
+  },
+  "abilities": {
+    "goal_framing": "目标定义",
+    "context_packaging": "上下文打包",
+    "agent_delegation": "Agent 调度",
+    "review_judgment": "评审判断",
+    "verification_discipline": "验证纪律",
+    "failure_recovery": "故障恢复",
+    "scope_control": "范围控制",
+    "shipping_hygiene": "交付卫生",
+    "product_judgment": "产品判断",
+    "collaboration_handoff": "协作交接"
+  }
+}

package/package.json

@@ -0,0 +1,36 @@
+{
+  "name": "@iiwish/agentrecord",
+  "version": "0.0.1",
+  "description": "Local-first, auditable AI work profiles from agent traces.",
+  "type": "module",
+  "bin": {
+    "agentrecord": "src/cli.mjs"
+  },
+  "files": [
+    "src",
+    "examples",
+    "locales",
+    "references",
+    "schemas",
+    "scripts",
+    "README.md"
+  ],
+  "scripts": {
+    "check": "node --check src/cli.mjs && node src/cli.mjs --help",
+    "pack:dry": "npm pack --dry-run --json",
+    "smoke:install": "node scripts/smoke-install.mjs"
+  },
+  "engines": {
+    "node": ">=20"
+  },
+  "keywords": [
+    "ai-agent",
+    "codex",
+    "claude-code",
+    "opencode",
+    "local-first",
+    "work-profile",
+    "audit-log"
+  ],
+  "license": "MIT"
+}

package/references/evidence-rules.json

@@ -0,0 +1,112 @@
+{
+  "schema_version": "agentrecord.evidence-rules.v0",
+  "scope": "generic",
+  "rules": [
+    {
+      "id": "EV-GENERIC-RELEASE-OPS",
+      "title": "Release and shipping verification discipline",
+      "category": "shipping_hygiene",
+      "evidence_level": ["E1", "E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["release", "published", "package", "registry", "CI", "tag", "version", "SHA", "deploy", "build verification"],
+        "exclude_any": ["wishlist", "someday"]
+      },
+      "signal_template": "Release and shipping work treats version, artifact, branch, package, deploy, or registry evidence as verifiable facts.",
+      "dimensions": ["Verification Discipline", "Shipping Hygiene", "Scope Control"],
+      "role_impacts": ["Technical Reviewer", "Shipping Owner", "Systems Thinker"],
+      "profile_impact": "Positive signal for turning AI-assisted work into reviewable release or delivery evidence instead of unsupported completion claims."
+    },
+    {
+      "id": "EV-GENERIC-EVIDENCE-GOVERNANCE",
+      "title": "Evidence hierarchy and audit-boundary discipline",
+      "category": "evidence_governance",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["evidence", "audit", "trace", "E1", "E2", "E3", "E4", "self-report", "review decision", "acceptance gate"],
+        "exclude_any": []
+      },
+      "signal_template": "Work separates system evidence, reproducible traces, summaries, self-report, review decisions, and acceptance boundaries.",
+      "dimensions": ["Review Judgment", "Verification Discipline", "Agent Delegation", "Collaboration Handoff"],
+      "role_impacts": ["Technical Reviewer", "Agent Operator", "Shipping Owner"],
+      "profile_impact": "Positive signal for making AI-agent output inspectable, accountable, and safe to hand off."
+    },
+    {
+      "id": "EV-GENERIC-RELIABILITY-RECOVERY",
+      "title": "Reliability diagnosis and failure recovery",
+      "category": "reliability_recovery",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["debug", "failure", "crash", "timeout", "locked", "recovery", "retry", "smoke", "benchmark", "diagnosis", "regression"],
+        "exclude_any": []
+      },
+      "signal_template": "Reliability work uses staged diagnosis, failure isolation, smoke checks, benchmarks, retries, or recovery evidence.",
+      "dimensions": ["Failure Recovery", "Verification Discipline", "Review Judgment", "Scope Control"],
+      "role_impacts": ["Technical Reviewer", "Systems Thinker", "Shipping Owner"],
+      "profile_impact": "Positive signal for recovering from complex failures without broad rewrites or false success claims."
+    },
+    {
+      "id": "EV-GENERIC-PRODUCT-VALIDATION",
+      "title": "Product validation and launch-readiness judgment",
+      "category": "product_validation",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["MVP", "trial", "launch", "onboarding", "user validation", "viability", "positioning", "beta", "acceptance", "readiness"],
+        "exclude_any": []
+      },
+      "signal_template": "Product judgment ties scope, readiness, positioning, or trial decisions to visible validation evidence.",
+      "dimensions": ["Product Judgment", "Verification Discipline", "Goal Framing", "Scope Control"],
+      "role_impacts": ["Product Builder", "Shipping Owner", "Technical Reviewer"],
+      "profile_impact": "Positive signal for deciding when AI-assisted work is ready for users, trials, launch, or further narrowing."
+    },
+    {
+      "id": "EV-GENERIC-AGENT-WORKFLOW",
+      "title": "AI-agent delegation and review workflow design",
+      "category": "agent_workflow",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["agent", "delegation", "handoff", "execution packet", "prompt", "worker", "review flow", "task graph", "next prompt"],
+        "exclude_any": []
+      },
+      "signal_template": "AI-agent workflow work covers delegation, handoff, execution packets, review loops, or worker coordination.",
+      "dimensions": ["Agent Delegation", "Context Packaging", "Collaboration Handoff", "Goal Framing"],
+      "role_impacts": ["Agent Operator", "Shipping Owner", "Collaboration Handoff"],
+      "profile_impact": "Positive signal for operating agents as reviewable collaborators rather than one-off chat tools."
+    },
+    {
+      "id": "EV-GENERIC-PLATFORM-CONTRACT",
+      "title": "Platform contract and schema discipline",
+      "category": "platform_contract",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["contract", "schema", "protocol", "API", "router", "validation", "type", "migration", "compatibility"],
+        "exclude_any": []
+      },
+      "signal_template": "Platform work aligns contracts, schemas, protocols, APIs, migrations, validation, or compatibility behavior.",
+      "dimensions": ["Context Packaging", "Shipping Hygiene", "Collaboration Handoff", "Verification Discipline"],
+      "role_impacts": ["Systems Thinker", "Shipping Owner", "Technical Reviewer"],
+      "profile_impact": "Positive signal for protocol-minded engineering and cross-surface behavior alignment."
+    },
+    {
+      "id": "EV-ACTIVITY-METADATA",
+      "title": "Codex activity metadata baseline",
+      "category": "agent_usage",
+      "evidence_level": ["E2"],
+      "match": {
+        "include_all": [],
+        "include_any": [],
+        "exclude_any": []
+      },
+      "fallback": true,
+      "signal_template": "Local Codex session metadata establishes the activity baseline when no curated memory rule matches are available.",
+      "dimensions": ["Agent Delegation", "Context Packaging"],
+      "role_impacts": ["Agent Operator"],
+      "profile_impact": "Useful baseline signal for repeated agent usage, but insufficient for strong ability claims without stronger work evidence."
+    }
+  ]
+}

package/references/evidence-rules.zh-CN.json

@@ -0,0 +1,112 @@
+{
+  "schema_version": "agentrecord.evidence-rules.v0",
+  "scope": "generic",
+  "rules": [
+    {
+      "id": "EV-GENERIC-RELEASE-OPS",
+      "title": "发布与构建验证存证",
+      "category": "shipping_hygiene",
+      "evidence_level": ["E1", "E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["release", "published", "package", "registry", "CI", "tag", "version", "SHA", "deploy", "build verification"],
+        "exclude_any": ["wishlist", "someday"]
+      },
+      "signal_template": "发布与交付工作将版本、制品、分支、包、部署或注册表作为可审计验证的事实证据。",
+      "dimensions": ["Verification Discipline", "Shipping Hygiene", "Scope Control"],
+      "role_impacts": ["Technical Reviewer", "Shipping Owner", "Systems Thinker"],
+      "profile_impact": "高价值信号：表明能将 AI 辅助的工作输出，转化为有事实审计链保护的制品、分支与上线记录，而非口头上的“已写完”。"
+    },
+    {
+      "id": "EV-GENERIC-EVIDENCE-GOVERNANCE",
+      "title": "数据边界与审计合规存证",
+      "category": "evidence_governance",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["evidence", "audit", "trace", "E1", "E2", "E3", "E4", "self-report", "review decision", "acceptance gate"],
+        "exclude_any": []
+      },
+      "signal_template": "工作流程中清晰地划分系统事实级证据、可复现的追踪审计、汇总报告、自我陈述以及接收验收红线。",
+      "dimensions": ["Review Judgment", "Verification Discipline", "Agent Delegation", "Collaboration Handoff"],
+      "role_impacts": ["Technical Reviewer", "Agent Operator", "Shipping Owner"],
+      "profile_impact": "高价值信号：表明拥有清晰的主客体责任意识，使 AI 调度具备可审计、可追溯的安全交付合规属性。"
+    },
+    {
+      "id": "EV-GENERIC-RELIABILITY-RECOVERY",
+      "title": "高可靠性诊断与故障恢复存证",
+      "category": "reliability_recovery",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["debug", "failure", "crash", "timeout", "locked", "recovery", "retry", "smoke", "benchmark", "diagnosis", "regression"],
+        "exclude_any": []
+      },
+      "signal_template": "系统故障调试工作使用阶段式精准诊断、故障源头隔离、冒烟测试、基准基准测试、熔断重试等事实证据保障系统可靠性。",
+      "dimensions": ["Failure Recovery", "Verification Discipline", "Review Judgment", "Scope Control"],
+      "role_impacts": ["Technical Reviewer", "Systems Thinker", "Shipping Owner"],
+      "profile_impact": "高价值信号：表明在面对高难度异常、死锁和编译崩溃时，能用分步求证法配合测试验证精准修复，而非无头苍蝇式乱改。"
+    },
+    {
+      "id": "EV-GENERIC-PRODUCT-VALIDATION",
+      "title": "产品可行性验证与就绪判定存证",
+      "category": "product_validation",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["MVP", "trial", "launch", "onboarding", "user validation", "viability", "positioning", "beta", "acceptance", "readiness"],
+        "exclude_any": []
+      },
+      "signal_template": "产品就绪判定和业务落地深度绑定，将产品功能边界、灰度上线及用户可用性决策与公开事实证据链牢牢挂钩。",
+      "dimensions": ["Product Judgment", "Verification Discipline", "Goal Framing", "Scope Control"],
+      "role_impacts": ["Product Builder", "Shipping Owner", "Technical Reviewer"],
+      "profile_impact": "高价值信号：表明具有出色的业务敏感度，懂得何时叫 AI 快速停止无谓的过度设计，专注核心 MVP 验证。"
+    },
+    {
+      "id": "EV-GENERIC-AGENT-WORKFLOW",
+      "title": "智能体协同调度与工作流设计存证",
+      "category": "agent_workflow",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["agent", "delegation", "handoff", "execution packet", "prompt", "worker", "review flow", "task graph", "next prompt"],
+        "exclude_any": []
+      },
+      "signal_template": "AI 协同操作工作涵盖了任务分派、优雅交接、上下文数据打包、多轮评审反馈闭环以及多智能体协同调度。",
+      "dimensions": ["Agent Delegation", "Context Packaging", "Collaboration Handoff", "Goal Framing"],
+      "role_impacts": ["Agent Operator", "Shipping Owner", "Collaboration Handoff"],
+      "profile_impact": "高价值信号：表明不再将 AI 当作一锤子买卖的问答聊天框，而是将其作为可以被规范管理、调度并参与工程评审的协同下属。"
+    },
+    {
+      "id": "EV-GENERIC-PLATFORM-CONTRACT",
+      "title": "契约模式与 API 架构规约存证",
+      "category": "platform_contract",
+      "evidence_level": ["E2", "E3"],
+      "match": {
+        "include_all": [],
+        "include_any": ["contract", "schema", "protocol", "API", "router", "validation", "type", "migration", "compatibility"],
+        "exclude_any": []
+      },
+      "signal_template": "架构底层工作对准并对齐核心契约、数据 Schema、网络协议、API 路由规约、版本平滑迁移以及兼容性判定。",
+      "dimensions": ["Context Packaging", "Shipping Hygiene", "Collaboration Handoff", "Verification Discipline"],
+      "role_impacts": ["Systems Thinker", "Shipping Owner", "Technical Reviewer"],
+      "profile_impact": "高价值信号：表明具有敏锐的架构治理视野，擅长使用确定性的 Schema 和接口契约去约束 AI 从而保证大型系统的稳定性。"
+    },
+    {
+      "id": "EV-ACTIVITY-METADATA",
+      "title": "Codex 本地会话活动度基线",
+      "category": "agent_usage",
+      "evidence_level": ["E2"],
+      "match": {
+        "include_all": [],
+        "include_any": [],
+        "exclude_any": []
+      },
+      "fallback": true,
+      "signal_template": "在缺乏结构化调教文本证据时，使用本地 Codex 适配器累计的原始会话数据和 Token 活动度建立基础行为基线。",
+      "dimensions": ["Agent Delegation", "Context Packaging"],
+      "role_impacts": ["Agent Operator"],
+      "profile_impact": "基础级存证信号：能作为持续、高频使用 AI 进行本地开发的基本活动度证明，但在缺乏上层事实证据时，不直接代表工程交付水平。"
+    }
+  ]
+}

package/schemas/evidence.schema.json

@@ -0,0 +1,30 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://agentrecord.local/schemas/evidence.schema.json",
+  "title": "AgentRecord Evidence Card",
+  "type": "object",
+  "required": [
+    "id",
+    "level",
+    "title",
+    "summary",
+    "agent_clients",
+    "dimensions",
+    "role_signals",
+    "confidence",
+    "refs",
+    "privacy"
+  ],
+  "properties": {
+    "id": { "type": "string" },
+    "level": { "type": "array", "items": { "type": "string" } },
+    "title": { "type": "string" },
+    "summary": { "type": "string" },
+    "agent_clients": { "type": "array", "items": { "type": "string" } },
+    "dimensions": { "type": "array", "items": { "type": "string" } },
+    "role_signals": { "type": "array", "items": { "type": "string" } },
+    "confidence": { "type": "string" },
+    "refs": { "type": "array" },
+    "privacy": { "type": "object", "required": ["public_safe"] }
+  }
+}

package/schemas/profile.schema.json

@@ -0,0 +1,34 @@
+{
+  "$schema": "https://json-schema.org/draft/2020-12/schema",
+  "$id": "https://agentrecord.local/schemas/profile.schema.json",
+  "title": "AgentRecord Profile",
+  "type": "object",
+  "required": [
+    "schema_version",
+    "owner",
+    "generated_at",
+    "report",
+    "work_identity",
+    "work_role_signals",
+    "ability_model",
+    "agent_ledger",
+    "evidence_notes",
+    "calibration_notes",
+    "privacy_boundary",
+    "run_metadata"
+  ],
+  "properties": {
+    "schema_version": { "const": "agentrecord.profile.v0" },
+    "owner": { "type": "object", "required": ["id", "display_name"] },
+    "generated_at": { "type": "string" },
+    "report": { "type": "object", "required": ["locale", "audiences", "schema_language"] },
+    "work_identity": { "type": "object", "required": ["primary_label", "summary", "confidence", "evidence_ids"] },
+    "work_role_signals": { "type": "array", "items": { "type": "object", "required": ["role_id", "label", "band", "confidence", "evidence_ids"] } },
+    "ability_model": { "type": "array", "items": { "type": "object", "required": ["dimension_id", "label", "score", "band", "confidence", "evidence_ids"] } },
+    "agent_ledger": { "type": "object", "required": ["clients"] },
+    "evidence_notes": { "type": "array", "items": { "type": "object", "required": ["evidence_id", "level", "title", "summary", "refs", "privacy"] } },
+    "calibration_notes": { "type": "array" },
+    "privacy_boundary": { "type": "object", "required": ["local_only_generation", "raw_logs_included", "public_session_ids_included"] },
+    "run_metadata": { "type": "object", "required": ["mode", "run_count", "generated_at"] }
+  }
+}