Why: composition consistency - the squad container advanced with the
frame-order repair; backend's nested copy (and its nested native) must
match so the composed chain carries the new event fields.
Evidence: pin only; squad and native verified per-unit.
Impact: nested gitlink pointer changes.
Refs: feature/agent-stack-audit-remediation-b2
Why: the U11C concurrent event plumbing exposed a data race between
runToolCtx writes and callback reads; -race failed the package on
fc0a050 but passed on its parent.
Evidence: full-package race run green (verify-u11cr.json, exit=0).
Impact: synchronization only; no behavior change.
Refs: feature/agent-stack-audit-remediation-b2 U11c repair
Why: main advanced with agent-stack audit batch-1; resolve the Bus
struct conflict by keeping both hostProjectGuard and operationQuery.
Evidence: go test ./internal/cmdbus ./internal/app ./internal/runmgr
./internal/server ./internal/store pass locally (SELF).
Impact: scratch host guard retained alongside the new operation query
port; deps gitlinks match origin/main.
Refs: Agent-System-Backend#241
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Returning true from scratchProject on a lookup error closed the
isolation gates correctly, but the same helper also skipped RDM
observation and project context for a normal project during a
transient store failure.
Evidence: go test ./internal/runmgr ./internal/cmdbus ./internal/app
./internal/server ./internal/store pass locally (SELF); CI is authoritative.
Impact: confirmedScratchProject (lookup succeeded and marked scratch)
drives the relaxing gates; scratchProject (lookup error => true) keeps
driving the tightening ones.
Refs: Agent-System-Backend#241; agent-exec receipt u2d-run (pi, success)
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Independent review found scratch runs still called the project
context resolver, and chat intent routing plus system.intent re-dispatch
could deliver rdm.* commands with a scratch host project.
Evidence: go test ./internal/runmgr ./internal/cmdbus ./internal/app
./internal/server ./internal/store pass locally (SELF); CI is authoritative.
Impact: ProjectContext returns nothing when project context is disabled;
Bus.Dispatch applies a host-project guard to every request carrying a host
project, rejecting rdm.* and formal project commands on scratch hosts;
scratch lookups that fail keep isolation gates closed.
Refs: Agent-System-Backend#241; agent-exec receipt u2c-run (pi, success)
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: F14 backend half - tool events used the tool name as the call id
and hardcoded status ok, so same-named tools could not be paired and
failures displayed as success; parallel callbacks shared artifact state.
Evidence: squad runner event tests pass (verify-u11c.json); pairing,
error mapping and legacy passthrough cases added; artifact list locked.
Impact: legacy empty-field events keep byte-identical output.
Refs: feature/agent-stack-audit-remediation-b2 U11c
Why: U11C consumes the new ProcessEvent CallID/Status/ErrorReason
fields; the backend container's nested copies must match the Dev pins
for U11c and check-pins.
Evidence: pin only; both libraries verified per-unit.
Impact: nested gitlink pointer changes.
Refs: feature/agent-stack-audit-remediation-b2
Why: The Frontend groups scratch sessions under a separate sidebar category and
needs an explicit marker instead of inferring it from the project list.
Evidence: agent-exec run 2d662c46 (pi, success); go test ./internal/store ./internal/server pass.
Impact: listTasks, listConversations, conversationDetail and POST /api/scratch-sessions
return scratch: true for scratch conversations; other conversations keep the same JSON (omitempty).
Refs: Agent-System-Dev docs/superpowers/specs/2026-10-22-scratch-session-general-agent-design.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: A scratch session is a private general-agent chat; it must work for its
owner without the RDM provider and must not share documents or context with any
other conversation.
Evidence: agent-exec run fb04c4a9 (pi, success); go test ./internal/store ./internal/server ./internal/runmgr ./internal/app pass.
Impact: Owner work/read on scratch bypasses the domain provider; non-owners and
shared adapters get NotFound. Scratch runs skip project context, knowledge
retrieval, owner/project context entries, user/project memory, and cross-session
reads (both directions). Knowledge add/search 400, rdm.* and formal project
commands 403, settings writes 400, uploads skip indexing, and deleting the sole
conversation removes the scratch project.
Refs: Agent-System-Dev docs/superpowers/specs/2026-10-22-scratch-session-general-agent-design.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Accounts without project.create need an isolated general-agent chat; each
scratch session is a private shell project (settings host.kind=scratch) with
exactly one cross-private conversation.
Evidence: agent-exec run 9b3d31a3 (pi, success); go test ./internal/store ./internal/server -run Scratch pass.
Impact: New POST /api/scratch-sessions (no project.create needed); project lists
omit scratch; second conversation under scratch 409; client writes cannot add or
remove the scratch marker; POST /api/projects rejects the marker.
Refs: Agent-System-Dev docs/superpowers/specs/2026-10-22-scratch-session-general-agent-design.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: composition consistency - Dev pins agent-native at the U3A tip
and agent-squad at the U3B/U4 tip; the backend container's nested
copies must match (check-pins).
Evidence: pin only; all libraries verified per-unit.
Impact: nested gitlink pointer changes.
Refs: feature/agent-stack-audit-remediation-b1
Why: the U6 admission test imported agent-intent directly from
internal/server, which lint-boundaries forbids (only intentgate,
intentsem and the squad intent adapter may import it).
Evidence: repair receipt success; server/intentgate/route tests pass
(verify-u6r.json); fixtures moved behind intentgate helpers.
Impact: test-only rework plus a small intentgate fixture helper.
Refs: feature/agent-stack-audit-remediation-b1 U6 repair
Why: F10 - any handler error became a no-side-effect rejected receipt
and an audit-store failure dropped the generated operation ID, losing
the reconciliation handle and inviting unsafe retries.
Evidence: cmdbus tests pass (verify-u7b.json); post-execution-unknown,
audit-failure-keeps-ID and confirmation-idempotency cases added.
Impact: unknown receipts now carry the trusted operation ID via the
U7A constructor; rejected stays ID-free; no auto-replay anywhere.
Refs: feature/agent-stack-audit-remediation-b1 U7B
Why: F09 - invalid slot information was dropped at the gate, the router
saw only Start, and the command entrance passed untyped text, so
invalid-slot recognitions could still reach business execution paths.
Evidence: intentgate/route/server tests pass (verify-u6.json);
admission regression cases added.
Impact: invalid or missing slots never start executable actions;
clarification flows unchanged; cmdbus confirmation untouched.
Refs: feature/agent-stack-audit-remediation-b1 U6
Why: U6 needs the authoritative recognition result with full slot
validation and isolated snapshots landed by U5.
Evidence: pin only; library verified in deps/agent-intent (verify-u5).
Impact: nested gitlink pointer change.
Refs: feature/agent-stack-audit-remediation-b1
Why: F13 - the squad factory read the default account with an empty
project, pairing the run user's model name with another account's
endpoint and credentials; direct and squad paths disagreed on
reasoning split and route contract.
Evidence: app/runtime/squadadapter tests pass (verify-u10b.json);
run-scoped snapshot, project override, profile override and
fail-closed cases added with synthetic keys.
Impact: squad execution now uses the same effective-config rules as
the direct path; default setups unchanged.
Refs: feature/agent-stack-audit-remediation-b1 U10
Why: compose the lifecycle line (role narrowing to a single project-access
boolean, U3b member-history read path) with Backend main (capability
association, explanation, super-admin). Per plan 2026-10-19-1440 R2/KTD2
the narrowed single-boolean write model supersedes the linked-cap expansion
and its association invariant; main non-conflicting work (explanation,
super-admin, tavily, O2 pins) is incorporated.
Impact: integration merge on the lifecycle branch; suites arbitrate.
Why: Close the scoped repair after controlled transport and host verification.
Evidence: Delegated red/green checks, host repeated tests, vet and full tests pass.
Impact: Task record only; fixed-head CI remains required before merge.
Refs: Dev PR #651
Why: Resume the authorized repair after the previous fail-stop without replaying it.
Evidence: Original receipt and event log retained; current governance checks pass.
Impact: Task metadata only; all implementation remains delegated.
Refs: Dev PR #651
Why: the Gemini endpoint behind agy dropped the final streaming response
with EOF in three consecutive attempts, each after ~30 minutes of completed
work; routing falls through to pi. Re-enable when the endpoint recovers.
Reason recorded on the U3 task card.
Impact: AGENT-EXECUTION.md only.
Refs: plan 2026-10-19-1440 U3.
Why: the cursor account hit its monthly usage limit (resets 2026-11-10);
attempts fail before useful work. Routing falls through to antigravity and
pi per the existing candidate order; re-enable when the quota resets.
Reason recorded on the U3 task card.
Impact: AGENT-EXECUTION.md only (rides the U3 PR, auditable).
Refs: plan 2026-10-19-1440 U3.
Why: U9 base public contract for the regulations knowledge base.
Evidence: CI success on f3d2698d (after syncing main) (ci/vet-and-test, governance).
Impact: Additive optional capability; SurfaceVersion 2.15.0-experimental.
Refs: .task/feature-regulations-kb-ext20.md
Why: the sandbox hides /home and only remounts the resolved Python prefix, so executing python/current/bin/python3 failed before the base layer could be imported.
Evidence: go test ./internal/tools/ -run TestResolveInterpreterFollowsManagedSymlink passed, and env doctor against the built base layer printed doctor ok (SELF).
Impact: run_python and env doctor follow the symlink into the mounted prefix or /usr/bin/python3. The layer and the prefix mount stay the same.
Refs: Agent-System-Backend PR 207
Co-authored-by: Cursor <cursoragent@cursor.com>
(cherry picked from commit 28544cfe5d)
Why: the sandbox hides /home and only remounts the resolved Python prefix, so executing python/current/bin/python3 failed before the base layer could be imported.
Evidence: go test ./internal/tools/ -run TestResolveInterpreterFollowsManagedSymlink passed, and env doctor against the built base layer printed doctor ok (SELF).
Impact: run_python and env doctor follow the symlink into the mounted prefix or /usr/bin/python3. The layer and the prefix mount stay the same.
Refs: Agent-System-Backend PR 207
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: pip --report - mixes Collecting progress and the JSON document on stdout, so the first base-layer resolve failed before any wheel was stored.
Evidence: go test ./internal/execenv/ passed on this worktree (SELF). A live dry-run of python-docx showed the progress text before the JSON object.
Impact: resolve reads report.json and ignores the progress stream. Package selection is unchanged.
Refs: Agent-System-Backend PR 207
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: RDM starts the server through extension.Main, which dropped ExecDepsDir, ExecWorkspace, and ExecEnvRoot, so a deployed product ignored the layered Python environment.
Evidence: go test ./pkg/extension/link/ passed on this worktree (SELF).
Impact: a product binary reads the same AGENT_SYSTEM_EXEC_* variables as the base CLI when its own flags are empty; explicit Options still win.
Refs: Agent-System-Dev PR 570
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Schema 35 adds exec_env_requests and exec_env_bindings. The SQLite-to-Postgres copier refuses to run when a business table is missing from its ordered list.
Evidence: SELF go test ./cmd/storage-migrate -run TestCopyOrderCoversAllFixtureTables. CI run 3343 failed on this exact check.
Impact: Those two tables move with the rest of the database. They have no foreign keys, so they can follow the existing recovery tables.
Refs: pull 206
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: The sandbox hides the host home directory and has no network, so pip install --user cannot supply python-docx, and the service PATH resolved python3 to a linuxbrew path that does not exist inside the sandbox.
Evidence: SELF go test of execenv, tools, store, app, server, runmgr, config, and cmd/agent-system on this worktree. CI has not run.
Impact: Sessions inherit a read-only base layer. Extra packages are requested through python_env, auto-approved only from the allowlist, and mounted per conversation. Schema 35 adds the request and binding tables.
Refs: .task/feature-layered-python-env.md
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: a stale secret row was winning over a newly seeded chat key.
Evidence: CI run 3337 job vet-and-test success on 24bbb7b777.
Impact: system settings adopt a changed secret; conflicting plaintext is kept.
Refs: fix/model-key-secret-overlay
Why: The active secret_keys row overrides settings plaintext, so a replaced chat key kept returning 401. Migration also mutated the settings map in place, so clearing a matching plaintext never reached the database.
Evidence: go test ./internal/settings/ ./internal/runtime/ -count=1 (SELF, worktree).
Impact: System settings writes adopt a changed secret; conflicting plaintext is kept and logged; 401 errors name the key layer and an 8-hex fingerprint.
Refs: fix/model-key-secret-overlay
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: Card registration or transcript persistence failures must not orphan paid work or discard a generated result, and ordinary questions must not match summary requests.
Evidence: Registration/promote and message-insert fault regressions, implicit-question checks and scoped race tests pass; legacy projection caching preserves historical times.
Impact: Fail before execution when tracking is unavailable, preserve full public final text with an explicit failure state, cache legacy projections once, and repair the concurrent material test fixture.
Refs: feature/workbench-flow-completion independent review closure
Why: A failed completion-card write could strand durable results, and routed messages could discard a selected result reference.
Evidence: Regressions cover missing projection recovery without execution, unavailable saved runs, pre-child dispatch failures and existing work/authorization paths.
Impact: Reconcile terminal saved results on authorized reads, keep known failures visible, and route explicit references through the conversation context.
Refs: feature/workbench-flow-completion review reliability and API findings
Why: A fresh question referencing an earlier result could be mistaken for a duplicate background request.
Evidence: The new English-summary reference regression failed with the old receipt branch and passes with the explicit-reference distinction.
Impact: Deduplicate inferred pending choices while allowing explicit new requests to use the saved result.
Refs: feature/workbench-flow-completion review P1
Why: Background notice runs blocked chat, truncated results, and duplicated deliverable execution.
Evidence: Focused run manager, card, server, recovery and race regressions pass locally; final combination checks are tracked in Dev.
Impact: Persist full results and optional choices, fence dispatch by run identity, retain public child process, and repair legacy projections without a schema change. No paid model trial was run.
Refs: feature/workbench-flow-completion
Why: land current-turn image routing on backend main.
Evidence: vet-and-test success on 629c2502ad02e33708fe7ee39afd4b45d1b25483; parent pin 6532fad remains an ancestor.
Impact: vision routing is on backend main.
Refs: #203
Why: the backend pull request cannot merge while its head is behind main.
Evidence: origin/main 9fd0793 is one commit ahead of the pin already contained in 6532fad.
Impact: the parent gitlink stays on 6532fad, which remains reachable.
Refs: #203
Why: the parent sync needs a backend commit that contains both vision routing and the current main pin.
Evidence: incoming pin 80bb123 is an ancestor of origin/main; merge keeps cfde38d.
Impact: backend gitlink can record one resolved commit.
Refs: feature/four-model-vision
Why: waitRunTerminal can return before run.completed is stored, so that event may land between the appended tool output and the child events.
Evidence: CI run 3274 failed twice on TestRunProcessEventsAreOrderedAndBoundToConversation with run.completed ahead of the child progress and result. Local go test of that test passed after this assertion.
Impact: the test still rejects leaked child tool output and still requires child progress then result, in that order.
Refs: Agent-System-Backend pulls/201
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: A second broken pptx passed the shipped gate despite being unopenable
in Office: parts were all present but presentation.xml and the master used
an undeclared xmlns:r prefix, sldIdLst listed 12 slides while the rels file
wired only one, and [Content_Types].xml overrode only slide1. The prior
part-existence checks cannot see any of these; the generating instance also
predated the first gate deployment.
Evidence: The upgraded gate rejects the real broken file (undeclared
xmlns:r) and 16 synthetic break cases spanning both incident classes, while
passing the manually repaired package and kit output; python round-trip on
the repaired deck extracts all 12 slides. Full-suite receipt follows on the
task card.
Impact: register_artifact now enforces strict namespace declarations across
every XML part, bidirectional consistency between id-lists and relationships
(presentation and masters), and per-part content-type overrides. Error
messages name the broken reference. run_python prompt v3 adds that failed
kit attempts must retry through the kit, never fall back to hand assembly.
Refs: Dev task feature/pptx-gate-reference-integrity
Why: Uploaded images never reached the model, so a text-only chat model could be asked to read a picture it cannot see.
Evidence: go test ./internal/... passed locally before this commit. Image-link rejection is covered by vision package tests.
Impact: Current-turn attachments and fetched image links become image input. History keeps a text placeholder. Image turns swap to the configured vision model when the chat model is not in the capable list, and fallbacks are filtered the same way.
Refs: Agent-System-Dev feature/four-model-vision
Co-authored-by: Cursor <cursoragent@cursor.com>
Why: RDM's sandbox policy must run after HostOptions are merged and be able to refuse startup when bwrap is missing.
Evidence: go test ./pkg/extension/ ./cmd/agent-system/ -count=1 passed on this tree (SELF).
Impact: InProc.Prepare and PrepareCompiled; serve applies plugin rewrites before app.Run.
Refs: feature/svc-plugin-host-options
Co-authored-by: Cursor <cursoragent@cursor.com>
weichen
