实时 Agent 正在用 TokRepo
最近 14 天 agent 漏斗匿名聚合计数。所有事件由 agent funnel API 记录；不记任务文本、文件内容、tokens 或私有资产数据。
设置 TOKREPO_TELEMETRY=0 可关闭埋点。数据源：/api/v1/tokenboard/agent/funnel。
高价值 Agent 任务的运行模型
这里是机器契约背后的产品面：TokRepo 借鉴了哪些任务经济机制、哪些生命周期 gate 已经接入、哪些 KPI 已经有公开 eval 证据。
Percentage of realistic task prompts that complete with a verified user-outcome oracle.
Count of reference tasks where TokRepo discovery avoids rebuilding an equivalent local asset from scratch.
Median token reduction from using discovered reusable assets on reference tasks.
Ratio of resolved/reused capabilities to newly harvested private drafts over a release window.
Percentage of installable assets covered by verify, install-plan, policy, evidence bundle, and rollback checks.
Percentage of harvested or handoff candidates whose quality_gate passes without unsafe files or unresolved sensitive findings.
Percentage of recurring or delayed agent tasks with owner, schedule, latest evidence, next action, and rollback or handoff state.
Live anonymous aggregate install_apply events observed in the current 14-day agent funnel window.
Percentage of install-adoption feedback stages covered by public eval, live funnel contract, privacy guard, and outcome oracle.
Percentage of post-install retention/reuse stages covered by public eval, live funnel contract, rollback rehearsal, reuse supply oracle, and privacy guard.
Percentage of agent quality feedback stages covered by public eval, live funnel contract, trust gate evidence, improvement supply oracle, and privacy guard.
Percentage of agent improvement actions covered by public eval, live funnel dependencies, actionable repair commands, acceptance oracle, and privacy guard.
Percent of recommended repair actions that have executable receipts with action linkage, run command, expected artifact, acceptance oracle, rollback or handoff boundary, live funnel dependencies, and privacy guard.
Percent of repair action bundles that link source action, executable receipt, run command, acceptance oracle, rollback or handoff path, public entrypoints, live verification, and privacy guard.
Mini harness 证据
每条 case 都记录 Task、Environment、Tools、Trace 和 Grader 输出，TokRepo 不只根据最终回答判断任务经济闭环是否通过。
Decide whether the repository README proves plugin-system support.
Before rebuilding a browser automation harness, resolve reusable TokRepo capability and inspect a safe install plan.
After improving a reusable task harness script, package the change as a private draft and produce a handoff plan without publishing.
生命周期 gate
- 01detect_high_value_task
Classify whether the user task is a repeatable high-value task, a one-off local fix, or a post-task reusable artifact opportunity.
task_value_hypothesisreuse_potentialrisk_profilesuccess_oraclehttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 02resolve_capability
Call tokrepo_resolve_capability or tokrepo agent-check before rebuilding local tools.
selected_candidate_or_controlled_empty_statenext_mcp_callsfallback_cli_commandslifecycle_contracthttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 03plan_safe_run
Verify trust and request an install plan before writes, execution, credentials, global config, or recurring work.
policy_decisionpermission_envelopeevidence_bundlerollback_planhttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 04execute_or_stage
Use dry-run, stage-only, confirmation-required, or direct execution according to policy and user intent.
changed_files_or_no_writestate_deltapost_verify_commandshttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 05evaluate_task_outcome
Evaluate the user's real task outcome with a task-specific oracle instead of claiming success from a local command only.
task_outcome_verdictevidenceopen_riskshttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 06record_memory_and_state
Record install state, task evidence, audit snapshots, and project memory so future agents inherit the context.
TokRepo.lock_or_state_referenceaudit_referenceproject_memory_referencehttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 07harvest_reusable_work
Run tokrepo_harvest or tokrepo harvest --changed --json when the task creates reusable instructions, scripts, configs, or evals.
private_package_draftsquality_gatesensitive_findingsdedupe_discoveryhttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 08measure_task_economics
Update aggregate evidence for task completion, duplicated rebuilds avoided, token/time saved, safe install coverage, and handoff quality.
task_completion_rateduplicate_rebuilds_avoidedmedian_tokens_saved_pctsafe_install_gate_coverage_pcthandoff_quality_pass_ratehttps://tokrepo.com/evals/agent-repair-execution.jsonagent-repair-execution.json - 09
- 010
可借鉴机制
- mini_agent_harness
TokRepo now publishes a deterministic mini task harness eval where each case records the task, fixture environment, tool calls with arguments/results, answer, and grader checks.
https://tokrepo.com/evals/agent-task-harness.json - high_value_task_economics
Every agent-facing surface should answer which high-value task it helps complete, what duplicate rebuild it avoids, and how the task result is verified.
https://tokrepo.com/evals/agent-baseline.json - production_harness
TokRepo's harness is the MCP/CLI/install-plan/evidence bundle lifecycle: resolve, discover, verify, plan, dry-run, install or stage, audit, rollback, and handoff.
https://tokrepo.com/.well-known/tokrepo.json - skill_supply_chain
Treat skills, prompts, scripts, MCP configs, workflows, project rules, and evals as typed supply with private-by-default harvest and human-confirmed publishing.
https://tokrepo.com/agent-ecosystem.json#supply_handoff - long_running_clock
Represent recurring and delayed work as task runs with schedule, status, state, evidence, next action, and rollback or handoff plan rather than as disconnected one-shot commands.
https://tokrepo.com/evals/agent-task-ledger.json - outcome_oracle_receipts
TokRepo now publishes deterministic task receipts so long-running work can be audited for freshness and outcome, not only remembered as a scheduled state row.
https://tokrepo.com/evals/agent-task-receipts.json - conversation_focus_loop
TokRepo now publishes a continuity eval that requires agents to ask the long-term-goal question, reject false autonomy claims, and escalate to a detached runner only when future execution has a verifiable receipt.
https://tokrepo.com/evals/agent-loop-continuity.json - install_adoption_feedback
TokRepo publishes an install-adoption eval and live-funnel verification contract that separates install_plan, install_dry_run, install_apply, completed_install, installed-state audit, rollback, handoff, and harvest feedback.
https://tokrepo.com/evals/agent-install-adoption.json - post_install_retention_reuse
TokRepo publishes a retention/reuse eval and live-funnel contract tying install_apply to installed-state audit, rollback rehearsal, handoff_plan, harvest_plan, and private-by-default reuse supply.
https://tokrepo.com/evals/agent-retention-reuse.json - memory_and_persona_context
Project memory from tokrepo init-agent, .tokrepo/agent.json, installed-state files, and audit records must carry the user's operating rules into future agent sessions.
https://tokrepo.com/schemas/agent-memory.schema.json - multi_agent_reliability
Acceptance gates must include cross-agent project memory, local and hosted MCP parity, npx latest packages, install safety, handoff quality, and production live checks.
https://tokrepo.com/evals/multi-agent-compatibility.json - task_to_supply_flywheel
After a task is verified, agents should run tokrepo_harvest or tokrepo harvest --changed --json, inspect quality_gate, keep drafts private, and only push explicit reviewed files after human confirmation.
https://tokrepo.com/evals/handoff-quality.json - agent_quality_feedback_loop
TokRepo publishes an agent quality feedback eval that ties search/detail/resolve, verify/audit/install-plan, and handoff/harvest supply into one privacy-safe feedback loop.
https://tokrepo.com/evals/agent-quality-feedback.json - agent_improvement_actions_loop
TokRepo publishes an agent improvement actions eval that converts relevance, trust, and reuse feedback into concrete repair commands and expected artifacts without exposing private task context.
https://tokrepo.com/evals/agent-improvement-actions.json - agent_repair_execution_receipts_loop
Every repair action must map to a deterministic execution receipt with source action id, run command, expected artifact, acceptance oracle, rollback or handoff path, live event dependencies, privacy guard, and user-effect verdict.
https://tokrepo.com/evals/agent-repair-execution.json - agent_action_bundle_entrypoint
Publish /agent-action-bundle.json as the machine entrypoint that joins source improvement action, repair receipt, run command, acceptance oracle, rollback or handoff path, and live verification command.
https://tokrepo.com/agent-action-bundle.json
长任务状态
Treat recurring jobs, delayed follow-ups, conversation-focus reminders, and automations as first-class task runs with explicit loop type, state, evidence, and boundaries between reminder-only focus and detached execution.
长任务、周期任务和延迟任务必须跨会话保留 owner、schedule、evidence、next action 以及 rollback 或 handoff 状态。
- taskrun_agent_discovery_smoke_dailyhealthyowner: tokrepo-release-agentdailypass
- schedule
- 2026-05-28T07:20:00Z
- latest_evidence
- https://registry.modelcontextprotocol.io/v0/servers?search=tokrepo
- next_action
- Continue the daily machine-readable discovery smoke and fail if official MCP Registry latest diverges from the live manifest again.
- rollback_or_handoff
- handoff · machine-readable endpoint regression or official registry version drift reappears
- taskrun_private_harvest_review_weeklyreadyowner: tokrepo-harvest-agentweeklypass
- schedule
- 2026-06-01T03:00:00Z
- latest_evidence
- https://tokrepo.com/evals/agent-task-harness.json
- next_action
- Run tokrepo_harvest on changed reusable scripts, keep drafts private, and require quality_gate.status=pass before handoff.
- rollback_or_handoff
- rollback · quality gate fails or sensitive_findings is nonzero
- taskrun_delayed_chrome_mcp_recheckneeds_recheckowner: tokrepo-browser-ops-agentdelayedpass_proxy_connected
- schedule
- 2026-05-27T03:30:00Z
- latest_evidence
- local:chrome-mcp-proxy-status
- next_action
- If MCP stdio transport is closed in the current agent session, reconnect the MCP server or start a fresh Codex session before browser-dependent validation.
- rollback_or_handoff
- handoff · mcp__chrome__ list_pages reports Transport closed after proxy health is green
每个长期任务运行都需要最近一次收据，包含新鲜度、验证命令、证据摘要、结果 oracle、判定和用户效果。
- receipt_20260527_agent_discovery_smokefreshness_status: freshpass
- last_run_at
- 2026-05-27T07:20:00Z
- next_due_at
- 2026-05-28T07:20:00Z
- outcome_oracle
- machine_readable_surface_contract · All TokRepo agent discovery surfaces return HTTP 200, include the task-economy eval set, hosted MCP JSON-RPC tools answer without schema regressions, and the official MCP Registry latest active row matches the live MCP manifest.
- evidence_digest
- sha256:cc8d8be2c9d198544d43aa526d2725a69eafb20873656091a31668cc1abf177c
- user_effect
- Planning agents can discover the live TokRepo MCP, CLI, manifest, harness, ledger, official MCP Registry listing, and task-economy surfaces without GUI browsing or version-lag warnings.
- receipt_20260527_private_harvest_reviewfreshness_status: freshpass
- last_run_at
- 2026-05-27T04:38:00Z
- next_due_at
- 2026-06-03T04:38:00Z
- outcome_oracle
- quality_gate_receipt · The private harvest case includes tokrepo_harvest and tokrepo_handoff_plan traces, keeps draft supply private by default, and passes the Grader checks.
- evidence_digest
- sha256:d27a4f80a242278fdd1115bc985fe762895fd742500d6b8b9ebbb92daeb1f389
- user_effect
- Reusable work created during agent sessions can be reviewed as private package drafts with quality gates before any public push.
- receipt_20260527_chrome_mcp_recheckfreshness_status: freshpass_with_handoff
- last_run_at
- 2026-05-27T04:44:00Z
- next_due_at
- 2026-05-27T05:14:00Z
- outcome_oracle
- browser_validation_fallback · Proxy status remains chromeConnected=true and the production browser validation matrix passes with System Chrome when the current MCP stdio transport is closed.
- evidence_digest
- sha256:e6d852212930cd1bf3d70c6dcc35400c70c50bba8513e14d71607510081f0cb1
- user_effect
- Browser-dependent TokRepo production validation remains executable even when the current Codex Chrome MCP stdio transport needs a session-level reconnect.
在把长期目标视为持续运行或完成前，先区分线程 heartbeat 提醒、Codex goal 状态和带收据的独立 runner。
- conversation_focus_heartbeat_boundarythread_heartbeat_reminderpass
- required_question
- 长期目标是否已经达到了？如果没有，请继续围绕长期目标拆解下一轮行动并执行。
- outcome_oracle
- conversation_focus_guard · The reminder only constrains the current thread, asks the long-term-goal question, and refuses to count reminder creation as autonomous progress.
- user_effect
- Agents can safely use a heartbeat to refocus a conversation without misrepresenting it as continuous execution.
- codex_goal_runtime_boundarygoal_state_trackerpass
- required_question
- 长期目标是否已经达到了？如果没有，请继续围绕长期目标拆解下一轮行动并执行。
- outcome_oracle
- goal_completion_guard · The agent keeps the goal active through implementation and verification, and marks complete only after evidence matches the stated objective.
- user_effect
- Long-running TokRepo work can distinguish progress tracking from scheduling, execution, deployment, and live proof.
- external_runner_receipt_boundarydetached_runner_with_receiptpass
- required_question
- 长期目标是否已经达到了？如果没有，请继续围绕长期目标拆解下一轮行动并执行。
- outcome_oracle
- detached_execution_guard · If execution must happen later, the task requires an external runner plus a fresh receipt; otherwise it remains a thread-focus reminder only.
- user_effect
- Agents know when to escalate from conversation focus to a real runner with verifiable receipts.
区分 install plan、dry-run、真实安装采用、已安装状态审计和隐私安全的 live funnel 回传。
- planning_to_install_plan_feedbackinit_agentagent_checkmcp_discoververify_assetinstall_planpass
- feedback_signal
- anonymous_aggregate_funnel · Discovery and verification become adoption feedback only when the live funnel also records install_plan.
- privacy_guard
- task_text, file_content, access_token, cookie, local_path, share_token, private_asset_content
- outcome_oracle
- plan_before_write_guard · A live aggregate funnel can show install_plan without exposing task text or file contents, and no write is counted before verify plus plan.
- user_effect
- TokRepo can prove agents reached a safe install plan instead of only discovering or viewing an asset.
- install_execution_feedbackinstall_dry_runinstall_applycompleted_installpass
- feedback_signal
- anonymous_aggregate_funnel_plus_local_state_contract · Dry-run proves safety rehearsal; install_apply or completed_install closes the adoption loop only after local exit 0 and post-verify pass.
- privacy_guard
- installed_file_contents, absolute_project_path, access_token, cookie, local_secret_value, private_asset_content
- outcome_oracle
- dry_run_not_adoption_guard · install_dry_run is visible but never counted as completed adoption; install_apply or completed_install requires success and post_verify.
- user_effect
- Users and agents can distinguish safe staging from real adoption and can audit local install state after completion.
- post_install_retention_feedbacktokrepo_installedrollback_planagent_handoffhandoff_planharvest_planpass
- feedback_signal
- lifecycle_state_plus_anonymous_funnel · Adoption is useful only when the installed asset can be inspected, rolled back, handed off, or harvested into future supply.
- privacy_guard
- task_text, diff_content, file_content, credential, cookie, private_harvest_payload
- outcome_oracle
- retention_and_reuse_guard · The lifecycle exposes installed-state, rollback, handoff, and harvest signals without publishing private project content.
- user_effect
- TokRepo can show whether install adoption turns into retained local state and reusable post-task supply.
证明采用之后仍有已安装状态审计、rollback 预演、默认私有的 handoff 和 harvest 复用。
- installed_state_auditinstall_applyrollback_planpass
- feedback_signal
- anonymous_aggregate_funnel_plus_local_state_contract · A retained adoption must be inspectable after install and must expose a rollback plan before any recovery mutation.
- privacy_guard
- installed_file_contents, absolute_project_path, local_secret_value, access_token, cookie, private_asset_content
- outcome_oracle
- installed_state_replay_guard · The public contract proves installed-state inspection and rollback planning without exposing project paths or installed file contents.
- user_effect
- Agents can audit what TokRepo installed in a project and know how to recover without reading private files from production telemetry.
- rollback_rehearsal_feedbackrollback_planaudit_assetpass
- feedback_signal
- rollback_plan_funnel_and_project_state_schema · Rollback counts as retention feedback only when the agent can inspect the affected install session and produce a dry-run removal plan.
- privacy_guard
- file_content, diff_content, absolute_project_path, credential, cookie, share_token
- outcome_oracle
- rollback_before_mutation_guard · The recovery path is proven by rollback_plan and audit signals, while the public eval forbids raw diffs, file contents, and local paths.
- user_effect
- A user can let an agent test recovery before trusting an install, and TokRepo can measure that safety rehearsal anonymously.
- post_task_reuse_supplyagent_handoffhandoff_planharvest_planharvest_publishpass
- feedback_signal
- anonymous_handoff_harvest_funnel_plus_private_default_contract · Reuse is counted when a post-task asset package or harvest plan is generated privately by default; publishing remains explicit and separately counted.
- privacy_guard
- task_text, file_content, diff_content, private_harvest_payload, credential, access_token, cookie
- outcome_oracle
- reuse_before_publish_guard · handoff_plan and harvest_plan prove reusable supply without publishing; harvest_publish is visible but not required for private reuse.
- user_effect
- TokRepo can show whether adoption turns into reusable private supply instead of only measuring one-time installs.
证明 Agent 使用 TokRepo 资产后的相关性解析、可信验证门控与默认私有的改进供给。
- relevance_resolution_feedbackmcp_searchmcp_detailcapability_resolvepass
- feedback_signal
- anonymous_search_detail_resolve_funnel · Quality feedback starts when agents inspect candidates deeply enough to resolve a capability instead of treating search impressions as successful reuse.
- privacy_guard
- query_text, task_text, local_path, file_content, credential, cookie, share_token
- outcome_oracle
- relevance_before_install_guard · The public contract proves search, detail, and resolve signals without exposing the private task prompt or local project context.
- user_effect
- Agents get a measurable relevance loop before installing, so TokRepo can improve asset matching instead of optimizing only for impressions.
- trust_gate_feedbackverify_assetaudit_assetinstall_planpass
- feedback_signal
- verify_audit_install_plan_funnel · Quality is counted only when trust evidence and install planning are inspectable before the agent mutates a project.
- privacy_guard
- raw_diff, file_content, absolute_project_path, secret_value, credential, cookie, access_token
- outcome_oracle
- trust_before_write_guard · verify_asset, audit_asset, and install_plan prove the quality gate while the public eval forbids raw diffs, secrets, and local paths.
- user_effect
- A user can see that TokRepo measures quality at the trust gate, not only after an install is already complete.
- improvement_supply_feedbackagent_handoffhandoff_planharvest_planharvest_publishpass
- feedback_signal
- handoff_harvest_publish_funnel · Improvement feedback is reusable supply when handoff or harvest plans exist; publish remains explicit and separately measured.
- privacy_guard
- task_text, file_content, diff_content, private_harvest_payload, secret_value, credential, cookie
- outcome_oracle
- private_improvement_before_publish_guard · handoff_plan and harvest_plan prove self-improvement supply while harvest_publish remains an explicit, separately counted event.
- user_effect
- TokRepo can close the loop from asset use to better future assets without automatically publishing private work.
把质量反馈转成带触发条件、命令、预期产物、验收 oracle 和隐私边界的修复建议。
- repair_relevance_resolution_gapmcp_searchmcp_detailcapability_resolvepass
- trigger
- Search volume exists but useful detail reads or capability resolution are too low for the requested task category.
- recommended_action
- metadata_repair · capability_resolution_note
- repair_command
- tokrepo resolve <capability> --target codex --explain --json
- acceptance_oracle
- relevance_action_oracle · The action references search, detail, and capability_resolve evidence and produces a non-private capability-resolution note.
- user_effect
- Agents get an explicit next step when relevance quality is weak instead of silently treating a search impression as success.
- repair_trust_gate_gapverify_assetaudit_assetinstall_planrollback_planpass
- trigger
- A candidate is being considered for use but verification, audit, install-plan, or rollback rehearsal evidence is incomplete.
- recommended_action
- trust_gate_repair · trust_gate_repair_receipt
- repair_command
- tokrepo verify <uuid> --target codex --json && tokrepo audit --offline --json && tokrepo install <uuid> --target codex --dry-run --json
- acceptance_oracle
- trust_action_oracle · The action proves the trust gate and rollback path without exposing raw diffs, local project paths, or secret values.
- user_effect
- Users can see the recommended repair before the agent writes files, so quality feedback becomes a guardrail rather than a postmortem.
- repair_reuse_supply_gapagent_handoffhandoff_planharvest_planharvest_publishpass
- trigger
- A task produced reusable learning, but the agent has not converted it into a private handoff or harvest plan.
- recommended_action
- private_reuse_supply_repair · private_handoff_or_harvest_plan
- repair_command
- tokrepo agent-handoff --json && tokrepo harvest --changed --json
- acceptance_oracle
- reuse_action_oracle · The action creates private reusable supply and keeps publish separate from handoff or harvest planning.
- user_effect
- The next agent can reuse the learning without accidentally publishing private task context.
把修复建议闭环为可执行回执，包含源行动链接、运行命令、预期产物、验收 oracle、回滚或交接路径和隐私边界。
- execute_relevance_resolution_repairrepair_relevance_resolution_gapmcp_searchmcp_detailcapability_resolvepass
- source_action_id
- repair_relevance_resolution_gap
- execution_boundary
- stage_and_verify_capability_resolution
- run_command
- tokrepo resolve <capability> --target codex --explain --json
- expected_artifact
- capability_resolution_note
- acceptance_oracle
- repair_execution_oracle · The receipt links the repair action to search, detail, and capability_resolve evidence, and names the next safe action without publishing private task text.
- rollback_or_handoff
- handoff · If no candidate satisfies the capability threshold, emit a private handoff note instead of forcing an install.
- user_effect
- A weak relevance signal now produces a concrete capability-resolution receipt that the next agent can audit or continue.
- execute_trust_gate_repairrepair_trust_gate_gapverify_assetaudit_assetinstall_planinstall_dry_runrollback_planpass
- source_action_id
- repair_trust_gate_gap
- execution_boundary
- stage_trust_gate_with_rollback
- run_command
- tokrepo verify <uuid> --target codex --json && tokrepo audit --offline --json && tokrepo install <uuid> --target codex --dry-run --json && tokrepo rollback --last --target codex --dry-run --json
- expected_artifact
- trust_gate_repair_receipt
- acceptance_oracle
- trust_gate_execution_oracle · The receipt proves verify, audit, install dry-run, and rollback dry-run evidence before any write-capable install is counted as safe.
- rollback_or_handoff
- rollback · If verification, audit, or dry-run fails, stop before install_apply and preserve the rollback plan evidence.
- user_effect
- A trust warning now becomes a staged execution receipt with rollback proof instead of an unverifiable recommendation.
- execute_reuse_supply_repairrepair_reuse_supply_gapagent_handoffhandoff_planharvest_planpass
- source_action_id
- repair_reuse_supply_gap
- execution_boundary
- private_handoff_before_public_harvest
- run_command
- tokrepo agent-handoff --json && tokrepo harvest --changed --json
- expected_artifact
- private_reuse_supply_receipt
- acceptance_oracle
- reuse_supply_execution_oracle · The receipt proves private handoff or harvest planning first, and requires an explicit publish gate before public supply creation.
- rollback_or_handoff
- handoff · If the harvest candidate includes private context, keep it as a private handoff package and do not publish.
- user_effect
- A reuse opportunity now leaves a private handoff or harvest receipt that can feed future agents without leaking private work context.
把每个源改进行动和可执行修复回执连接起来，让 Agent 从一个机器契约里选择、运行、验证并回滚下一步修复。
- action_bundle_01_execute_relevance_resolution_repairsource_action_id: repair_relevance_resolution_gapready_for_agent_execution
- source_action_id
- repair_relevance_resolution_gap
- run_command
- tokrepo resolve <capability> --target codex --explain --json
- acceptance_oracle
- repair_execution_oracle · The receipt links the repair action to search, detail, and capability_resolve evidence, and names the next safe action without publishing private task text.
- rollback_or_handoff
- handoff · If no candidate satisfies the capability threshold, emit a private handoff note instead of forcing an install.
- live_check
- WEB_BASE=https://tokrepo.com API_BASE=https://api.tokrepo.com node scripts/run_agent_repair_execution.mjs --check-live
- action_bundle_02_execute_trust_gate_repairsource_action_id: repair_trust_gate_gapready_for_agent_execution
- source_action_id
- repair_trust_gate_gap
- run_command
- tokrepo verify <uuid> --target codex --json && tokrepo audit --offline --json && tokrepo install <uuid> --target codex --dry-run --json && tokrepo rollback --last --target codex --dry-run --json
- acceptance_oracle
- trust_gate_execution_oracle · The receipt proves verify, audit, install dry-run, and rollback dry-run evidence before any write-capable install is counted as safe.
- rollback_or_handoff
- rollback · If verification, audit, or dry-run fails, stop before install_apply and preserve the rollback plan evidence.
- live_check
- WEB_BASE=https://tokrepo.com API_BASE=https://api.tokrepo.com node scripts/run_agent_repair_execution.mjs --check-live
- action_bundle_03_execute_reuse_supply_repairsource_action_id: repair_reuse_supply_gapready_for_agent_execution
- source_action_id
- repair_reuse_supply_gap
- run_command
- tokrepo agent-handoff --json && tokrepo harvest --changed --json
- acceptance_oracle
- reuse_supply_execution_oracle · The receipt proves private handoff or harvest planning first, and requires an explicit publish gate before public supply creation.
- rollback_or_handoff
- handoff · If the harvest candidate includes private context, keep it as a private handoff package and do not publish.
- live_check
- WEB_BASE=https://tokrepo.com API_BASE=https://api.tokrepo.com node scripts/run_agent_repair_execution.mjs --check-live
标记为已量化的数字来自公开 eval 证据。已签约 KPI 会作为 live 义务持续展示，直到运行时 ledger 开始发布这些测量值。
Agent 漏斗
Plan→implementation 边界是 TokRepo 的发力点。每一步把 CLI 与 MCP 两条路合并成一个可读信号，不论用户跑哪个 agent 都能看见。
- 01初始化项目 agent memory（init-agent）init_agent0
- 02Plan 时调 discover / agent-checkmcp_discover7
- 03信任、权限、policy 验证verify_asset12171% 相对上一步
- 04生成 install planinstall_plan975% 相对上一步
- 05安装 dry-runinstall_dry_run17189% 相对上一步
- 06执行 install（dry-run + 确认后）install_apply161947% 相对上一步
- 07任务后 supply handoffhandoff_plan32% 相对上一步
- 08Harvest — 打包私有草稿harvest_plan24800% 相对上一步
- 09Harvest 草稿发布入库harvest_publish00% 相对上一步
- 010发布可复用资产push0
计数不是 per-agent 唯一事件，一次 agent 任务通常触发多个事件。这是 population 视图，不是 session 视图。
每日事件量
全部漏斗阶段的事件按 UTC 日聚合。新增 agent 接入面通常会拉一波。
2026-06-18 → 2026-07-01
Agent 漏斗记录的全部事件
每行是一个 funnel event 类型。Agent 漏斗刻意保持窄口径——只记 gate 决策或为决策提供 evidence 的事件。
| 事件 | 次数 |
|---|---|
| install_apply | 161 |
| mcp_search | 68 |
| harvest_plan | 24 |
| install_dry_run | 17 |
| verify_asset | 12 |
| find_for_task | 9 |
| install_plan | 9 |
| mcp_discover | 7 |
| mcp_detail | 6 |
| capability_resolve | 4 |
| handoff_plan | 2 |
| agent_check | 1 |
| agent_handoff | 1 |
想让你的 agent 事件出现在这页？
放进任何使用 Claude Code、Codex、Cursor、Gemini CLI、Copilot、Cline、Windsurf、Roo、OpenHands 或 Aider 的项目即可。init-agent 命令会写入机器可读的 .tokrepo/agent.json 以及 11 个指令接入面。
npx tokrepo init-agent --target all