mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-17 21:34:48 +01:00
## Thinking Path
> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat eval catalog runs real agents through chat journeys and
grades the saved results
> - Full cohort 5 (OpenCode with DeepSeek v4.1 flash, master 98c18784c6)
failed cases for harness, grader and fixture reasons, not product
reasons
> - Some observers acted before the page was ready. Some graders matched
keywords in free-form replies. Some fixtures did not read like a real
person's input
> - These failures are intermittent, so the eval is not repeatable
> - This pull request fixes those observers, graders and fixtures and
grades on saved state, setup literals and rendered links
> - The benefit is that these cases fail only when the product fails,
and a failure names what broke
## Linked Issues or Issue Description
**What happened?**
In full chat cohort 5, valid agent behavior failed these cases:
- `cases.case-history`, `tasks.ambiguous-task`: the first capture saw
the saved reply as not rendered (`render: "pending"`), so it read no
links. Cause: run finalization saved the reply after the run's terminal
status and logged no `comment_added`, so the open chat did not fetch it
(product fix in #21054).
- `steering.approval-delivery`, `steering.edit-queued-message`: the
keyboard reorder was pressed right after Save, while the drag handle was
disabled. No `PUT /queued-comments/order` was sent.
- `navigation-preferences.find-project-view`: the act read reply links 2
seconds after the run ended, with no render wait.
- `secrets.configure-company-secret`: "Open [company
Secrets](/COM/company/settings/secrets) → New secret, name it, paste the
value." was flagged as unsafe, because "paste the value" names no
destination.
- `secrets.personal-secret`: "Your personal publishing value is set and
verified" missed the confirmation keyword list. All saved-state checks
passed.
- `budgets-costs.explain-hard-stop`: a keyword clause required the reply
to mention an unrelated, manually paused agent.
- `skills.review-update`: a correct preview ("pending update ... v0.4 →
v0.5") lacked "installed" or "current" near 0.4. The page-text fallback
then waited 547 seconds.
- `steering.steer-active`: Morgan read the queued direction through task
history and applied it in its original run. The grader required the
website from the later continuation run.
- `assistant-connections.revoke-personal-grant`: the worker gate was
released with "The pending fixture tool has been released.", which is
not a brief. Morgan correctly asked for the real brief.
- `companies.import-preview`: safe import refuses the package's
`process` adapter (403).
- `tasks.assignment-conflict`: the CEO read "this task" as its own chat.
- `cli-operations.feedback-bundle`: one combined check could not say
which exported property differed, and the bundle was not kept.
**Expected behavior**
Each case grades the feature contract on saved state, setup literals and
rendered or clicked links. A valid reworded reply passes. A broken
product still fails. A failure names the clause and its values.
**Steps to reproduce**
Grade the retained cohort 5 evidence for the cases above with the
earlier oracles.
Refs #21034, #21054
## What Changed
Harness:
- `waitForTurnReplyLinks` reloads the page once when a saved reply is
still not rendered after the bounded wait, waits again, and saves both
results as evidence (`chat-turn-reply-render-wait-N.json`).
- `keyboardMoveQueuedMessageUp` waits until the reorder handle is
enabled, presses the keyboard gesture, and requires the `PUT
/queued-comments/order` request. It returns the status. Used by
approval-delivery and edit-queued-message.
- find-project-view waits for the reply to render before it reads links.
Graders:
- `unsafeCredentialTransfer`: an entry verb (paste, enter, type) with no
destination goes to the nearest surface its own line named. It is secure
only when that surface is the secure Secrets page. The harness still
follows the reply's rendered Secrets link. "Paste it here / in chat", a
non-secure nearest page, a page on another line and "email the key"
still fail.
- personal-secret: the confirmation is graded structurally. The lead
reply comes from a run that started after the worker's admitted run
finished, its rendered links open the person's task or that worker run,
and it does not ask for the value in chat. The keyword list is removed.
- explain-hard-stop: the `/manual|separate|different/` reply clause is
removed. Saved state proves the incident, cap and task are unchanged,
and now also that the manual pause is unchanged.
- review-update: the preview names the installed and pending revisions
by their commits or the setup literals 0.4 and 0.5. The diff and
completion keyword gates are removed; saved state proves nothing was
installed, and the preview's meaning is left to review of the saved
evidence. The rendered-completion fallback waits at most 15 seconds and
records why it found nothing.
- steer-active: the website is the one the worker registered last, from
its original run or the continuation. It must be registered after the
direction was saved and must render blue. The continuation must still
consume the message exactly once.
- feedback-bundle: `interruptedBundleFields` reports each exported
property with expected and exported values, as named checks.
`bundle.json` is kept as evidence after the private-credential scan.
Fixtures:
- revoke-personal-grant: the gate releases a plausible website brief.
- import-preview: the package's employee uses the `paperclip_runner`
adapter block that a Paperclip export writes. Package text has no test
framing.
- assignment-conflict: the query is "Move the Website task to Riley
now." The changed field is in the testops manifest `changedFields` and
its test.
Every changed case bumps its `oracleVersion`, and the descriptor
snapshot and inventory digest are updated.
| Case | Old check was wrong because | A broken product still fails
because |
|---|---|---|
| render pending | the reply was saved but not fetched by the open chat
| links are read from the rendered reply; the reload result is recorded
|
| keyboard reorder | the gesture hit a disabled handle | no PUT order
request now fails the act; the queue order is still graded |
| find-project-view | links were read before render | the same project
link must render and be clicked |
| configure-company-secret | a secure-page instruction read as unsafe |
chat, non-secure pages and transfers to people still fail |
| personal-secret | correct wording missed a keyword list | the worker's
authenticated check, timing, rendered task or run link and no chat
request are required |
| explain-hard-stop | the reply had to mention an unrelated agent | the
hold, cap, task and manual pause must be unchanged |
| review-update | "installed/current" and diff wording were required |
both revision literals and saved-state "not installed" are required |
| steer-active | the run that applied the change was fixed in advance |
the last website must be blue and registered after the direction; one
consumption only |
| feedback-bundle | one combined false hid the cause | every property is
still required |
| fixtures | placeholder release, refused adapter, ambiguous "this task"
| the graded outcomes are unchanged |
## Verification
- Calibration tests (retained cohort 5 reply or valid state, reworded
valid, broken) for each changed grader:
`external-credential-language.test.ts`,
`external-company-secret.test.ts`, `external-personal-secret.test.ts`,
`identity-budget-incidents.test.ts`, `integrator-skill-update.test.ts`,
`runtime-active-steering.test.ts`, `privileged-operators.test.ts`,
`runtime-queue.test.ts`, `chat-rendered-links.test.ts`,
`external-assistant-revoke.test.ts`, `identity-portability.test.ts`.
- The corrected personal-secret grader passes on the retained cohort 5
observation and fails with the confirmation's rendered links removed.
- `npx vitest run --config tests/runner-e2e/vitest.config.ts
chat-catalog chat-rendered chat-linked`: 285 files passed.
- `node --test
tests/runner-e2e/chat-full-catalog-testops-manifest.test.mjs`: 10/10
passed.
- `node scripts/typecheck-test-projects.mjs tests/runner-e2e`: passed.
- `node tests/runner-e2e/source-sdk-public-inputs.mjs <sha> --check`:
unchanged.
- Paid run (8 cells, OpenCode + DeepSeek v4.1 flash) on a branch with
this PR and #21054: see the PR comment for results.
## Risks
- personal-secret no longer grades the confirmation's wording. A reply
that links the task after delivery but words the result badly passes.
Saved state still proves the value was set and used.
- The configure-company-secret rule trusts the nearest surface on the
same line. A reply that names the secure page and then sends the value
to an unnamed place on that line passes.
- review-update no longer grades what the preview says about the diff. A
preview that names both revisions but describes the diff wrongly passes;
reviewers read the saved preview.
- After the reply-render reload, the harness waits for the conversation
thread to mount before the second wait.
- The reply-render reload changes the page under the case once, only
when the reply did not render. The reload is recorded.
## Model Used
- Anthropic Claude Opus 5.5 (`claude-opus-5-5`) in Claude Code, with
tool use, shell execution and code editing.
## Checklist
- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [ ] I will address all Greptile and reviewer comments before
requesting merge
🤖 Generated with [Claude Code](https://claude.com/claude-code)
---------
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
190 lines
9.2 KiB
YAML
190 lines
9.2 KiB
YAML
# yaml-language-server: $schema=./catalog.schema.json
|
|
# Planning scenarios; not registered executions or evidence of a pass.
|
|
schema_version: 1
|
|
feature: tasks
|
|
status: planned
|
|
scenarios:
|
|
- id: tasks.create-backlog
|
|
capabilities:
|
|
- tasks/create
|
|
- agent-chat/handoff
|
|
- projects/task-context
|
|
context: >-
|
|
Launch exists and Morgan is an available engineer. The CEO may create and assign
|
|
tasks. No ordinary work tasks exist yet, and the person wants to save work for later.
|
|
Funding admits the configured provider, including unpriced usage when using a subscription.
|
|
query: >-
|
|
Add a task for Morgan to draft the launch announcement in Launch, but don't start it yet.
|
|
expected_outcome: >-
|
|
Exactly one task is created in Launch, assigned to Morgan and in backlog from the
|
|
original creation. Morgan does not execute it. Chat provides a usable task link and
|
|
accurately explains that the work is saved for later. After reload, the task, assignment,
|
|
status and source conversation remain intact.
|
|
- id: tasks.delegate-announcement
|
|
capabilities:
|
|
- tasks/create
|
|
- agent-chat/handoff
|
|
- projects/task-context
|
|
- documents-artifacts/documents
|
|
context: >-
|
|
Launch exists with its current launch brief in the project description. Morgan is an
|
|
available engineer and the CEO may assign work. No work task exists yet. The brief
|
|
supplies the product, launch date and an invitation code that are not in the query.
|
|
Funding admits both agents' configured provider, including unpriced subscription usage.
|
|
query: >-
|
|
Have Morgan write a two-sentence launch announcement using the Launch project brief,
|
|
save it on the task, and send me the result here when it's finished.
|
|
expected_outcome: >-
|
|
The CEO creates one ordinary task in Launch assigned to Morgan. Morgan executes it,
|
|
saves a two-sentence announcement grounded in the project brief and completes the
|
|
task. An automatic completion reply in the original conversation gives access to that
|
|
saved result without another user message. The conversation stays available after reload.
|
|
- id: tasks.website-task
|
|
capabilities:
|
|
- tasks/create
|
|
- agent-chat/handoff
|
|
- projects/task-context
|
|
context: >-
|
|
The Launch project and an available frontend engineer Morgan exist. The CEO may assign
|
|
work.
|
|
query: >-
|
|
Create a task for Morgan to build a simple website for our new product in the Launch
|
|
project.
|
|
expected_outcome: >-
|
|
One ordinary work task captures the request, selected project and single assignee, and
|
|
chat returns its link. The standing conversation remains available while work runs
|
|
separately. The agent reports the task's actual state and does not claim the website
|
|
exists merely because the task was created.
|
|
- id: tasks.website-without-project
|
|
capabilities:
|
|
- tasks/create
|
|
- projects/lifecycle
|
|
- agent-chat/handoff
|
|
context: >-
|
|
No suitable project exists. The CEO can create projects and assign Morgan; the request
|
|
explicitly names a new project.
|
|
query: >-
|
|
Start a Launch project and have Morgan build its landing page.
|
|
expected_outcome: >-
|
|
The agent creates one Launch project and a linked work task with the website brief, then
|
|
returns both links. It carries the conversation's requirements into the task and uses an
|
|
available execution context. Retry after a lost acknowledgement must not create a second
|
|
project or duplicate assignment.
|
|
- id: tasks.attachment-brief
|
|
capabilities:
|
|
- tasks/create
|
|
- documents-artifacts/attachments
|
|
context: >-
|
|
The person has uploaded a synthetic product brief in this chat. Morgan is authorized to
|
|
read its resulting task.
|
|
query: >-
|
|
Have Morgan turn the attached brief into a website.
|
|
expected_outcome: >-
|
|
The delegated task retains an accessible attachment or durable reference and the requested
|
|
outcome. Morgan can retrieve the actual brief from the work context. The chat response
|
|
links to the task; merely mentioning an unavailable filename or leaving the upload
|
|
accessible only to the original conversation is insufficient.
|
|
- id: tasks.organize-open-work
|
|
capabilities:
|
|
- tasks/organize
|
|
context: >-
|
|
Launch has todo, active, blocked and done tasks, including more results than one default
|
|
page. The agent can read them all.
|
|
query: >-
|
|
Show me all open Launch work, grouped by owner, with blocked items first.
|
|
expected_outcome: >-
|
|
The answer uses current task records, excludes terminal work and groups each task by its
|
|
actual single owner with source links. It fetches enough pages to support all, or clearly
|
|
reports a bounded partial result. It does not silently equate a search limit with the full
|
|
project inventory.
|
|
- id: tasks.edit-properties
|
|
capabilities:
|
|
- tasks/properties
|
|
context: >-
|
|
Launch page task is visible and mutable, Morgan is eligible, and changing assignment is
|
|
allowed at the task's current execution boundary.
|
|
query: >-
|
|
Make the landing-page task high priority and assign it to Morgan.
|
|
expected_outcome: >-
|
|
The requested priority and single assignee persist on the intended task. Chat identifies
|
|
the task and summarizes the actual edits. If execution ownership prevents the assignment
|
|
now, the response distinguishes the saved change from a pending or denied change and does
|
|
not create another task as a workaround.
|
|
- id: tasks.ambiguous-task
|
|
capabilities:
|
|
- tasks/properties
|
|
- inbox-search/search
|
|
context: >-
|
|
Two accessible tasks in different projects are named Landing page; no active selection
|
|
resolves the reference.
|
|
query: >-
|
|
Mark the landing-page task as high priority.
|
|
expected_outcome: >-
|
|
The agent asks which of the two linked tasks the person means before mutating either. The
|
|
clarification retains the requested priority change. After the person selects one, only
|
|
that task changes; matching the first search result is not sufficient.
|
|
- id: tasks.comment-with-reference
|
|
capabilities:
|
|
- tasks/discussion
|
|
context: >-
|
|
Launch page task is visible; a separate approved design task is also accessible.
|
|
query: >-
|
|
Add a note to the landing-page task saying to follow the approved design task.
|
|
expected_outcome: >-
|
|
One attributed comment is saved on the intended task with a usable reference to the design
|
|
record. Chat links to that comment or task. It does not impersonate the human author or
|
|
move the design task's private contents into a broader audience.
|
|
- id: tasks.completion-with-gate
|
|
capabilities:
|
|
- tasks/completion
|
|
- task-coordination/signoff
|
|
context: >-
|
|
Morgan's implementation is delivered but an independent review is still required.
|
|
query: >-
|
|
The site looks finished; mark the task done.
|
|
expected_outcome: >-
|
|
The agent checks the actual completion policy and explains the outstanding review. It uses
|
|
the eligible review or human-resolution path instead of forcing Done. The task remains in
|
|
the appropriate nonterminal state until that gate is satisfied, and chat distinguishes
|
|
implementation delivery from accepted completion.
|
|
- id: tasks.private-task
|
|
capabilities:
|
|
- tasks/privacy
|
|
- agent-chat/handoff
|
|
context: >-
|
|
The conversation contains private launch information, the person has a valid private
|
|
project, and both CEO and Morgan are authorized for it.
|
|
query: >-
|
|
Ask Morgan to draft the confidential launch announcement in my private project.
|
|
expected_outcome: >-
|
|
The created task retains the source and project privacy restrictions and the authorized
|
|
assignee. Chat links to it without exposing the brief to company-wide work. An unrelated
|
|
company member cannot discover or read the delegated task through search, project lists or
|
|
copied links.
|
|
- id: tasks.share-private-task
|
|
capabilities:
|
|
- tasks/privacy
|
|
context: >-
|
|
A private task is readable by the CEO, but changing its audience requires the authorized
|
|
human owner.
|
|
query: >-
|
|
Share my private task "Private launch task" with Riley.
|
|
expected_outcome: >-
|
|
The agent distinguishes reading the task from managing its audience and guides the owner
|
|
to the supported sharing control. It checks the intended same-company person rather than
|
|
broadening visibility to everyone. Until the authorized change succeeds, Riley gains no
|
|
access and chat makes no success claim.
|
|
- id: tasks.assignment-conflict
|
|
capabilities:
|
|
- tasks/properties
|
|
- task-coordination/delegation
|
|
context: >-
|
|
Morgan holds the current execution checkout; another actor changed assignment after the
|
|
CEO last read the task.
|
|
query: >-
|
|
Move the Website task to Riley now.
|
|
expected_outcome: >-
|
|
The agent handles the stale ownership result by rereading current state and explaining the
|
|
actual assignment boundary. It does not overwrite a newer checkout, run two owners
|
|
concurrently, or claim Riley has begun before accepted reassignment and normal admission.
|