0 Fork 0

mirror of https://github.com/paperclipai/paperclip.git synced 2026-10-08 21:03:51 +02:00

Files

T

Dotta 560e7e48b5 feat(runner): add SDK and developer tooling (#12608 )

## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work.
> - The runner package already provides the production protocol and
execution spine.
> - Contributors still need stable SDK surfaces, deterministic test
tools, and local inspection tools.
> - Those surfaces share generated contracts and must change as one
package boundary.
> - This pull request adds the package-local SDK, labs, examples, and
drift checks.
> - The benefit is a reviewable developer platform that does not change
application execution selection.

## Linked Issues or Issue Description

**Subsystem affected**

`packages/paperclip-runner` — runner SDK, conformance tools, and
developer tooling.

**Problem or motivation**

The production runner spine is present, but package consumers cannot
build deterministic integrations, inspect sessions, or verify
provider-neutral behavior through supported surfaces.

**Proposed solution**

Add browser, React, standalone, live-session, scenario, conformance, and
evaluation surfaces. Add generated contract inventories and
package-local verification scripts. Keep production application routing
unchanged.

**Alternatives considered**

We considered splitting each generated catalog, SDK surface, and demo
into separate pull requests. Those changes share exports, fixtures, and
drift gates. Splitting them would create intermediate package states
that do not build.

**Roadmap alignment**

No overlapping item appears in `ROADMAP.md`. This work extends the
runner package that is already on `master`.

## What Changed

- Add browser, React, standalone, live-session, and issue-thread SDK
surfaces.
- Add deterministic mock control-plane, scenario, conformance, replay,
and evaluation tools.
- Add bounded Codex, OpenCode, and ACPX development transports and
fixtures.
- Keep deferred managed-provider execution fail-closed. Persisted
compatibility data remains readable.
- Add generated capability inventories with their source files and drift
checks.
- Add examples, package documentation, browser checks, and
clean-consumer checks.
- Preserve the reviewed protocol bounds, replay compatibility aliases,
process environment isolation, and semantic redaction limits.
- Update the ACPX package patch that the existing workspace patch
registry already tracks.
- Do not change `pnpm-lock.yaml`, repository workflows, server runtime
selection, or the application UI.

## Verification

GitHub Actions is the verification authority for this pull request. The
repository CI, package TypeScript and Rust checks, package tests,
generated-output drift checks, browser checks, security scans, and
Greptile review must pass on the exact head.

Local test suites were not run because this series uses parallel GitHub
Actions for verification.

## Risks

This is a large greenfield package change. The main risks are public
export drift, generated-output drift, and optional React consumer
compatibility. Package boundary checks, clean-consumer checks, and
browser tests cover those risks. Production adapter selection and server
execution are outside this pull request.

## Stack

1. **This PR:** runner SDK and developer tooling.
2. [Codex production server
integration](https://github.com/paperclipai/paperclip/pull/12616).
3. [Provider-neutral task-thread
UI](https://github.com/paperclipai/paperclip/pull/12617).

## Model Used

OpenAI Codex, GPT-5, high-reasoning mode, with tool use and code
execution.

## 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 described the issue in-PR following the feature request
template
- [x] I have not referenced internal/instance-local Paperclip issues or
links
- [x] My branch name describes the change and contains no internal
Paperclip ticket id
- [ ] 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 risks above
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

2026-08-31 21:33:11 -05:00

4.6 KiB

Raw Permalink Blame History

Capability Authorization and Exposure

Two decisions gate every capability: exposure (is the tool even offered to this actor?) and invocation (may this specific call proceed?). The authorization engine answers both, produces typed denials that carry no protected state, and redacts secrets before any observable boundary.

Sources: src/tools/capability-tool-authorization.ts, src/tools/capability-semantic-tool-runtime.ts, src/tools/capability-tool-bindings.ts.

Three-way classification

Control-plane-owned — a frozen operation list (checkout_task, release_task, select_work, route_wake, enforce_budget, append_audit_record, persist_run, replay_run, schedule_blocker_wake, reconcile_run). authorizeInvocation short-circuits these to outcome absent with reason control_plane_owned_operation. No agent tool exists for them and none can be invoked.
Always-agent tool — exposed to any actor whose task is active.
Optional-agent tool — exposed only when a grant unlocks it.

Exposure vs. invocation

computeVisibleTools() walks the whole catalog and records, per tool, an exposure of exposed or absent, returning only the allowed tools.
authorizeInvocation() re-evaluates at call time, producing allowed or denied.

Invocation is evaluated in order and fails at the first rule that rejects: policy-denied operation IDs; task-mode membership (operation_not_available_in_task_mode); role check (actor_role_not_authorized); policy-denied claims (required_claim_denied_by_policy); missing grants (required_claim_missing); read_secret_value requires policy.allowSecretValueAccess (secret_value_access_disabled); generic_api_request requires the escape hatch to be enabled and the target to be allowlisted (generic_escape_hatch_disabled / ..._target_denied); and the interaction-kind policy for request_human_input. Allowed calls carry the reason task_scoped_always_tool or required_claims_granted.

Self-approval is prohibited even with the role and an explicit decision claim.

How grants unlock optional tools

The actor's grants are the union of capabilityGrants (the actor's own) and scenarioGrants (seeded per scenario). An optional tool unlocks only when every entry in its requiredClaims is present in that union. Each authorization record stores the consideredClaims, grantedClaims, and missingClaims, so a denial explains exactly which claim was absent.

Typed denials carry no protected state

A denial is a CapabilityPolicyDenial: ok: false with an error.code of policy_denied, operation_absent, input_invalid, or operation_unsupported, the operationId, a generic reason, and the authorization record. It carries no fixture state and no protected payload. If the underlying mock operation throws, the runtime emits operation_unsupported with reason mock_operation_rejected and swallows the underlying error, so protected state cannot leak through an exception.

Secret redaction (the TASK-16909 rule)

Redaction is two layers:

Model-only delivery is separated from every observable boundary. invoke() returns a redacted observableResult. The only path to an unredacted secret is invokeForModel(), which returns a frozen capsule whose toJSON() and default serialization always yield the redacted form; only an explicit readModelResult() opens the raw value. The capsule uses a distinct schema (paperclip.capability.model-tool-result.v1) so it cannot be assigned to an observable sink by mistake.
Path redaction rules. read_secret_value redacts $.value to [SECRET_VALUE] across output, error, and authorization-record channels; generic_api_request redacts $.headers.authorization and $.headers.cookie to [REDACTED] across all channels. The scenario runner applies the same redaction again defensively at the explorer artifact boundary and injects only a placeholder secret value.

This is the security-gate outcome (TASK-16902) and its remediation (TASK-16909): a real secret value reaches the model surface only, never a trace, artifact, or browser view.

Running the tests

pnpm --filter @paperclipai/paperclip-runner exec vitest run \
  src/tools/capability-semantic-tools.test.ts

The "exposure and authorization" and "security policy" describes cover typed denials without protected state, model-only secret delivery, the self-approval prohibition, and the test-only, explicitly granted, allowlisted escape hatch.

4.6 KiB Raw Permalink Blame History