ZacharyZhang-NY/AgentCompany
1
0 Fork 0
Code Issues Pull Requests Actions Packages Projects Releases 2 Wiki Activity
1,360 Commits 46 Branches 7 Tags
075dea725a0d69cd7abb2db3bb4ca5c093fa02c4
Commit Graph
59 Commits
Author SHA1 Message Date
ZacharyZhang-NYandClaude Opus 5.5 6e210301db Q136: remove rollback records that marked restores nothing performed 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:53:26 -04:00
ZacharyZhang-NYandClaude Opus 5.5 f8f852c326 Q135: remove the thread replay that only digested events and reconstructed nothing 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:47:48 -04:00
ZacharyZhang-NYandClaude Opus 5.5 403c57785c Q134: a signal's category comes from its runtime source, so self-checks cannot pass as real traffic 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:43:21 -04:00
ZacharyZhang-NYandClaude Opus 5.5 3849b96192 Q133: remove chaos drills that never injected a fault, and the signals they wrote 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:39:36 -04:00
ZacharyZhang-NYandClaude Opus 5.5 e562444487 Q132: remove PR bot gates that no task completion or goal close consulted 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:36:07 -04:00
ZacharyZhang-NYandClaude Opus 5.5 fe45cd4d46 Q131: remove review agents and eval suites whose pass only counted configuration rows 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:33:03 -04:00
ZacharyZhang-NYandClaude Opus 5.5 074a1fc8e3 Q130: every open task says why it is not moving, when it resumes, and who must act 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:27:29 -04:00
ZacharyZhang-NYandClaude Opus 5.5 271a83a7bc Q129: a run with failed actions it never redid does not count as completed 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 21:15:47 -04:00
ZacharyZhang-NYandClaude Opus 5.5 6afd7bef2a Q128: every model call of a run is recorded with its purpose and listed in the run 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 20:55:18 -04:00
ZacharyZhang-NYandClaude Opus 5.5 1f484df894 Q127: erasure removes the subject's content and keeps only a receipt 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 20:42:23 -04:00
ZacharyZhang-NYandClaude Opus 5.5 11cdc4af20 Q125: imports stream through a bounded window and refuse files over 1 GiB 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 20:17:28 -04:00
ZacharyZhang-NYandClaude Opus 5.5 91fa6b5f1e Q123: files shared in a private conversation stay with its members 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 19:52:37 -04:00
ZacharyZhang-NYandClaude Opus 5.5 69b83d6b77 Q122: reading a recorded file checks its bytes against the recorded hash 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 19:44:51 -04:00
ZacharyZhang-NYandClaude Opus 5.5 8f192eabe7 Q121: a file id keeps the bytes it was recorded with; rewrites get new ids 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 19:34:44 -04:00
ZacharyZhang-NYandClaude Opus 5.5 cb1a4be167 Q117: commands run in an allowlisted environment, and records and approvals say host isolation is advisory 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 19:07:28 -04:00
ZacharyZhang-NYandClaude Opus 5.5 abfcbf4824 Q114: goal audits use the org's own auditor model and see pointers, not the executor's account 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 18:35:39 -04:00
ZacharyZhang-NYandClaude Opus 5.5 34c5557a32 Q108: deliverables under review are held until approved instead of written first 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 17:45:39 -04:00
ZacharyZhang-NYandClaude Opus 5.5 4765683b86 Q107: a task approval covers the content it was given; edits need a fresh one 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 17:34:40 -04:00
ZacharyZhang-NYandClaude Opus 5.5 558d678ba2 Q104: connector grants can narrow a toolkit to named actions and fixed argument values 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 17:05:23 -04:00
ZacharyZhang-NYandClaude Opus 5.5 4a6641fd49 Q101: a connector can hold several labelled company accounts, each granted and used by name 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 16:46:46 -04:00
ZacharyZhang-NYandClaude Opus 5.5 736c45c0ee Q100: reconnecting keeps the working connector account until the new one is authorized 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 16:32:46 -04:00
ZacharyZhang-NYandClaude Opus 5.5 6f8731cbac Q99: connector actions keep an execution record so a lost reply is not resent 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 16:23:23 -04:00
ZacharyZhang-NYandClaude Opus 5.5 f6447319a3 Q95: a run names the capabilities it could not load 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 15:55:33 -04:00
ZacharyZhang-NYandClaude Opus 5.5 7cd386a38c Q93: disconnect erases the secret here and revokes at the provider where it can 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 15:33:58 -04:00
ZacharyZhang-NYandClaude Opus 5.5 4668396f25 Q87: provider errors keep status and provider code; text never decides recovery 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 14:52:31 -04:00
ZacharyZhang-NYandClaude Opus 5.5 665bf3cae1 Q83: a model passes an agent check before it does agent work 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 14:19:48 -04:00
ZacharyZhang-NYandClaude Opus 5.5 d557e42676 Q77: a session keeps the model and account it started with 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 13:03:11 -04:00
ZacharyZhang-NYandClaude Opus 5.5 bb279ff85a Q70: provider cooldowns survive restarts 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 12:07:17 -04:00
ZacharyZhang-NYandClaude Opus 5.5 c6a0460035 Q69: provider failures classify by status and code; cooldowns follow Retry-After or back off 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 12:03:37 -04:00
ZacharyZhang-NYandClaude Opus 5.5 c2f8e656f6 Q65: every call path records real tokens, cache, reasoning and cost 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 11:24:54 -04:00
ZacharyZhang-NYandClaude Opus 5.5 168572dd76 Q63: model calls are priced and agent and department budgets limit real spend 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 11:08:38 -04:00
ZacharyZhang-NYandClaude Opus 5.5 46cd167c2c Q61: a failed status check no longer ends recovery of a background response 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 10:42:57 -04:00
ZacharyZhang-NYandClaude Opus 5.5 e809cfefea Q54: later sessions see what earlier ones already did and where they failed 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 09:35:27 -04:00
ZacharyZhang-NYandClaude Opus 5.5 ec2698575d Q53: stalled goal reviews leave a visible state and a resume path 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 09:25:46 -04:00
ZacharyZhang-NYandClaude Opus 5.5 ddc8b485e9 Q50: work moved to review gets a reviewer; review closes with the verdict 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 08:50:02 -04:00
ZacharyZhang-NYandClaude Opus 5.5 7659adbbdd Q49: a task that spends its run budget shows as stalled; edits start a fresh budget 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 08:38:02 -04:00
ZacharyZhang-NYandClaude Opus 5.5 7be072f8c5 Q40: tasks choose a writing mode; house style applies to conversational work only 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 07:28:47 -04:00
ZacharyZhang-NYandClaude Opus 5.5 645ce686bb Q39: runs load only model-compatible skills and state each skill's contract 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 07:22:08 -04:00
ZacharyZhang-NYandClaude Opus 5.5 4c475e5ceb Q30: documents keep every version 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 06:10:25 -04:00
ZacharyZhang-NYandClaude Opus 5.5 ec9d74d128 Q27: knowledge corrections keep a revision history 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 05:51:32 -04:00
ZacharyZhang-NYandClaude Opus 5.5 18128b5ad4 Q26: knowledge has scope, expiry, and supersession 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 05:47:03 -04:00
ZacharyZhang-NYandClaude Opus 5.5 b32e784469 Q19: pack the skills most relevant to the task first 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 04:48:44 -04:00
ZacharyZhang-NYandClaude Opus 5.5 f4d16eb46a Q17: budget every model request against the provider's context window 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 04:40:02 -04:00
ZacharyZhang-NYandClaude Opus 5.5 469f2f2c35 Q15: show which messages each agent actually read 
Runs record a context manifest once a model turn succeeds; the chat
marks each human message with the agents that read its current version.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 04:21:57 -04:00
ZacharyZhang-NYandClaude Opus 5.5 22632a4695 Q11: deleted messages and withdrawn commands never reach the model 
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 03:47:57 -04:00
ZacharyZhang-NYandClaude Opus 5.5 6570492daa Q3: record a structured handoff on every delegated task 
create_task now pins upstream files by content hash and carries the
source task, source thread, verified facts, open questions and
acceptance criteria into the assignee's prompt.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 02:35:12 -04:00
ZacharyZhang-NYandClaude Opus 5.5 56b0594f91 Move Composio, logo.dev and Antigravity keys into Settings 
The three third-party keys were hardcoded or read from the environment.
Operators now enter them in Settings > Integrations; they are stored
encrypted in the app-wide integration_keys table (migration 058).
An agent holding connector grants fails loudly when no Composio key is
saved. The goal-review test no longer depends on absolute dates.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
 2026-10-08 01:34:14 -04:00
ZacharyZhang-NY 327de38e78 Kimi Code joins as an OAuth plan provider: RFC 8628 device flow, verified against the live endpoints 
- New oauth/kimi-code-flow.ts mirrors the official kimi-cli: form-encoded device_authorization + token polling at auth.kimi.com with the required X-Msh-* device headers, pure poll interpreter, 15-minute tokens refreshed via refresh_token grant
- Agent loop runs chat completions at https://api.kimi.com/coding/v1 with the Bearer access token; default model kimi-for-coding (kimi-for-coding-highspeed for Allegretto+ plans)
- Live model listing via /coding/v1/models; credential refresh wired into the turn credential resolver
- Migration 057 seeds the provider (disabled) into existing organizations
- Wire protocol probed live: device code issued, authorization_pending poll observed, 401 envelope on the coding API confirmed
 2026-07-10 17:18:31 -04:00
ZacharyZhang-NY 813276de71 Kimi and Kimi CN join the provider catalog: OpenAI-compatible endpoints, verified live 
- kimi -> https://api.moonshot.ai/v1 (platform.kimi.ai keys), kimi-cn -> https://api.moonshot.cn/v1 (platform.kimi.com keys)
- Adapters via openAiCompatibleAdapter; agent loop via makeChatCompletionsTurn; live model listing
- Curated fallback models per current docs: kimi-k2.7-code, kimi-k2.7-code-highspeed, kimi-k2.6
- Migration 056 seeds both providers (disabled) into existing organizations
- Both endpoints probed live: OpenAI-style 401 error envelope confirmed
 2026-07-10 17:12:16 -04:00
ZacharyZhang-NY 26de3129b1 Goals end at achieved, not at an empty board: the goal-review loop 
The board draining to 280/280 done left the company idle under a
still-active goal, with 24 markdown files and zero code as its life's work.
Two structural gaps caused it: goals had no lifecycle (status was written
once as 'active' and never read again — nothing re-engages anyone when the
last task closes), and nothing anywhere distinguished plans from product,
so every "engineering" objective closed on documents.

The loop, closed:
- tasks.origin (migration 055) records provenance: chat_wake for mention
  vehicles (backfilled from their runs), goal_review for checkpoints, null
  for real work. Chatter and checkpoints can therefore never count as
  progress.
- spawnGoalReviews (each worker tick): an active goal with zero open tasks
  and real progress since the last review gets ONE review task for the goal
  owner. The review demands evidence — list the workspace, read the code,
  run the tests; documents are how work starts, not proof it happened — and
  must end in exactly one of: complete_goal, or the next wave of tasks with
  requires_filesystem engineering. A review closed with neither spawns
  nothing: termination is structural, not a counter.
- complete_goal (CEO or goal owner) finally lets someone say a goal is
  achieved, audited.
- CEO and department-head voices now state the doctrine: a goal that
  promises software closes on code that builds and passes tests in the
  workspace, and implementation is delegated with requires_filesystem: true.

On the live profile this arms one review: the active platform goal has 0
open tasks and 21 real completed ones, so the CEO gets the checkpoint on
next launch and the company starts building instead of stopping.
 2026-07-09 20:29:42 -04:00