fix(agent-ux): atomic claim --next --start, disclosed intent limit overrides, leaner list/search envelopes, MCP flat read parameters

.agents/pm/features/pm-f05lsg.toon

@@ -6,7 +6,7 @@ status: open
 priority: 1
 tags[4]: agent-ergonomics,ci-gates,discoverability,long-horizon
 created_at: "2026-08-01T13:09:26.104Z"
-updated_at: "2026-10-23T10:17:53.217Z"
+updated_at: "2026-10-31T21:50:08.744Z"
 author: "harness:claude-code"
 estimated_minutes: 720
 acceptance_criteria: "A replay corpus of wrong invocations is generated from the contract tables and covers every flag with a closed domain, every subcommand family, every required-argument omission and every mutually exclusive flag pair; Every declared error code is reachable from at least one corpus input, and a code with no reachable input is reported as such; A scorer reports the recovery-closure fraction and the list of unrecoverable refusals, scoring a lone nearest-match suggestion as not recoverable; The closure fraction is a ratchet in the gate registry that may not fall, with a negative control proving the gate can fail; The refusal envelope has a declared shape carrying the failing surface, the rejected value, the legal domain when known, and the exit code; Non-interactive init guidance warnings expose bounded executable inspection and explicit add commands while preserving skip/decline intent and user-authored documentation, with real isolated recovery and byte-preservation evidence"
@@ -47,7 +47,7 @@ dependencies[28]{id,kind,created_at,author,source_kind,author_source}:
   pm-gh1363,related,"2026-10-10T05:34:42.387Z","harness:codex","cli:update:dep",detected
   pm-998juj,discovered_from,"2026-10-10T11:46:33.219Z","harness:codex","cli:update:dep",detected
   pm-83cz0o,discovered_from,"2026-10-10T12:22:11.123Z","harness:codex","cli:update:dep",detected
-comments[86]:
+comments[87]:
   - created_at: "2026-08-18T19:18:11.981Z"
     author: "harness:codex"
     text: "Implementation start 2026-08-18: active in one coupled SDK-first command-contract tranche spanning canonical flag concepts and budgets, truthful invocation arity, complete closed-domain recovery, refusal-closure scoring, source-derived executable recovery obligations, and shared CLI/MCP/completion/docs discovery metadata. This item is claimed because its acceptance surface will be edited and verified in this branch; unrelated parents and backlog items remain open and unclaimed."
@@ -308,6 +308,9 @@ comments[86]:
   - created_at: "2026-10-23T10:17:53.217Z"
     author: "harness:codex"
     text: "Additional recovery corpus candidate discovered during pm-t2dqzijokl31: src/core/test/run-selectors.ts still advises adding a first test with --command, but the current test command contract refuses --command as unknown_option. Existing --add and --add-json alternatives are valid. Preserve this in the canonical recovery-closure programme rather than create a duplicate general UX item; current implementation remains scoped to linked-resource positional recovery and assignment-only evidence."
+  - created_at: "2026-10-31T21:50:08.744Z"
+    author: "harness:claude-code"
+    text: "Scope note from pm-gh1643 (GH-1643): a second refusal shape belongs in this closure gate. 'pm claim --next --start' told the caller to 'pass --next' while --next was in recovery.provided_fields (the same defect shape GH-1352 had on --if-available). pm-gh1643 fixes the claim-target refusals and adds a targeted spec. The class-wide invariant is still unenforced: no refusal's message, required text, next_steps or suggested_retry may instruct passing a flag already present in provided_fields. PmRefusalClosureObservation does not carry next_steps or provided_fields today, so the census cannot observe this; extend the observation and add a redundant_flag_advice finding, then run it over every grammar-derived probe."
 notes[3]:
   - created_at: "2026-10-04T16:59:01.232Z"
     author: "harness:codex"

.agents/pm/history/pm-7zs0.jsonl

@@ -579,3 +579,5 @@
 {"hash_algorithm":"sha256","ts":"2026-10-30T13:14:20.174Z","author":"harness:claude-code","author_source":"detected","agent_harness":"claude-code","agent_model":"claude-opus-5-5","agent_model_source":"probe","agent_instance":"7b0e58428f53e3aaeaafb372","agent_provenance":{"model":{"value":"claude-opus-5-5","source":"probe"},"effort":{"value":"xhigh","source":"environment"},"role":{"value":"implementer","source":"inferred","confidence":"medium","rule_version":"v2","evidence":["claim:pm-gh1626","claim:pm-gh1630","claim:pm-gh1633","lineage:pm-gh1630","lineage:pm-5t33or","lineage:pm-n7rr","lineage:pm-mpbb","lineage:pm-doxj"]},"topic":{"value":"workset:pm-gh1626+pm-gh1630+pm-gh1633","source":"inferred","confidence":"medium","rule_version":"v2","evidence":["claim:pm-gh1626","claim:pm-gh1630","claim:pm-gh1633","lineage:pm-gh1630","lineage:pm-5t33or","lineage:pm-n7rr","lineage:pm-mpbb","lineage:pm-doxj"]},"version":{"value":"2.1.312","source":"probe"}},"op":"comment_add","patch":[{"op":"add","path":"/metadata/comments/257","value":{"created_at":"2026-10-30T13:14:20.173Z","author":"harness:claude-code","text":"October 30 follow-up at 13:15 UTC: scheduled Auto Release run 41029596560 (08:54 UTC) ended cancelled after 15m13s, mid-pipeline. The workflow uses cancel-in-progress: false with a 180-minute job timeout, so this was an external cancellation, not a concurrency or timeout cut. It made no publication. Automation opened blocker issue #1632 for the earlier dispatcher failure 40999107873; it is now linked to the owner pm-gh1622, with no duplicate item. Live state: npm @unbrained/pm-cli is 2026.10.28, latest GitHub release is v2026.10.28, and there is no 2026.10.29 or 2026.10.30 cut, so no duplicate automatic cut can exist. Sentry has exactly one unresolved issue in 14 days: PM-CLI-3Q, 2 events, last seen 2026-10-29 03:11:51 UTC. Under the release gate's 14-day window it blocks every cut until it is resolved with verified pm-presets remediation; upstream PR170 is still open at 97d5d1bf4. Telemetry stack check at 13:13 UTC: all containers healthy, API synced at 13:09, queue drained, 74,676 events in 24h, newest event under one minute old. The last hour split 3,432 automation and 352 test source contexts, so test-run tagging from GH-1625 is visible in production rows."}},{"op":"replace","path":"/metadata/updated_at","value":"2026-10-30T13:14:20.174Z"}],"before_hash":"6b69ce088b398cc97ec761a311b01654b7dbff970bdeadcf17cf87c5ecda87b0","after_hash":"ed3825cf0237eaec733c5699030cfce4d351e709a0c08651e9a179fdb21eece6","item_hash_version":3,"context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"},"version":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"02a1320865bc15cf9c902dc6407444821147a86595014b0d2c30d2b4aabc12a5"}
 {"hash_algorithm":"sha256","ts":"2026-10-30T17:18:07.453Z","author":"harness:claude-code","author_source":"detected","agent_harness":"claude-code","agent_model":"claude-opus-5-5","agent_model_source":"probe","agent_instance":"7b0e58428f53e3aaeaafb372","agent_provenance":{"model":{"value":"claude-opus-5-5","source":"probe"},"effort":{"value":"xhigh","source":"environment"},"role":{"value":"implementer","source":"argv"},"topic":null,"version":{"value":"2.1.312","source":"probe"}},"op":"update","patch":[{"op":"replace","path":"/metadata/expected_result","value":"At October 31 06:00 UTC verify the next eligible daily cut against the latest merged delivery, including tag/npm identity, npm/Bun/npx/bunx consumers and fresh observability; preserve original reliability outcomes."},{"op":"replace","path":"/metadata/updated_at","value":"2026-10-30T17:18:07.453Z"}],"before_hash":"ed3825cf0237eaec733c5699030cfce4d351e709a0c08651e9a179fdb21eece6","after_hash":"726a9d13392bbff0d67d3ef6335573bd5af0e5d687a1d5d5ced38a40cb0a0a33","item_hash_version":3,"message":"Align the release-health expected result with the rearmed October 31 reminder","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"},"version":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"}}},"event_class":"maintenance","record_hash_version":1,"record_hash":"2251442c1d82111b7c9ac9c605d14f71ad1f1ad1f0855a6f25108160779225a1","reanchor_evidence":[{"before_hash":"6b69ce088b398cc97ec761a311b01654b7dbff970bdeadcf17cf87c5ecda87b0","after_hash":"eec31ea294980d7fe3bd9669dd82dc9502c2f2d67ce2bd31739e51cd9c9caf72","patch_hash":"8415e8b1b08ac08fb872b01e06a2c6322ade6d26de5a1cd4fda62d56a6a1cbc5","hash_algorithm":"sha256","item_hash_version":3,"record_hash_version":1,"record_hash":"ae50f6d1abbb2313aa5f6d3a23b0b8e80cd44b5b89a6662eda211d088af72c4f","record":{"hash_algorithm":"sha256","ts":"2026-10-30T17:18:07.453Z","author":"harness:claude-code","author_source":"detected","agent_harness":"claude-code","agent_model":"claude-opus-5-5","agent_model_source":"probe","agent_instance":"7b0e58428f53e3aaeaafb372","agent_provenance":{"model":{"value":"claude-opus-5-5","source":"probe"},"effort":{"value":"xhigh","source":"environment"},"role":{"value":"implementer","source":"argv"},"topic":null,"version":{"value":"2.1.312","source":"probe"}},"op":"update","patch":[{"op":"replace","path":"/metadata/expected_result","value":"At October 31 06:00 UTC verify the next eligible daily cut against the latest merged delivery, including tag/npm identity, npm/Bun/npx/bunx consumers and fresh observability; preserve original reliability outcomes."},{"op":"replace","path":"/metadata/updated_at","value":"2026-10-30T17:18:07.453Z"}],"before_hash":"6b69ce088b398cc97ec761a311b01654b7dbff970bdeadcf17cf87c5ecda87b0","after_hash":"eec31ea294980d7fe3bd9669dd82dc9502c2f2d67ce2bd31739e51cd9c9caf72","item_hash_version":3,"message":"Align the release-health expected result with the rearmed October 31 reminder","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"},"version":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"}}},"event_class":"maintenance","record_hash_version":1,"record_hash":"ae50f6d1abbb2313aa5f6d3a23b0b8e80cd44b5b89a6662eda211d088af72c4f"}}]}
 {"hash_algorithm":"sha256","ts":"2026-10-31T00:01:31.558Z","author":"harness:claude-code","op":"merge_reconcile","patch":[],"before_hash":"726a9d13392bbff0d67d3ef6335573bd5af0e5d687a1d5d5ced38a40cb0a0a33","after_hash":"726a9d13392bbff0d67d3ef6335573bd5af0e5d687a1d5d5ced38a40cb0a0a33","message":"record field-aware branch merge provenance","item_hash_version":3,"context":{"merge":{"receipts":[{"receipt_id":"fe024a5f-0084-4e67-882f-b85ebc7b88c8","item_id":"pm-7zs0","item_path":".agents/pm/reminders/pm-7zs0.toon","conflict_fields":[],"fields_from_theirs":["expected_result"],"union_fields":[],"requested_preference":"ours","requested_preference_applied":false,"conflict_resolution":"latest_document_update","decisions":[]}]}},"event_class":"substantive","record_hash_version":1,"record_hash":"1a1377a900f98fb21321bbfc1e4a41d5223b4f4f67cba66b208ef6ef487aab66"}
+{"hash_algorithm":"sha256","ts":"2026-10-31T22:12:47.452Z","author":"harness:claude-code","author_source":"detected","agent_harness":"claude-code","agent_model":"claude-opus-5-5","agent_model_source":"probe","agent_instance":"8a4268fe12c2a5f9fcd4da8d","agent_provenance":{"model":{"value":"claude-opus-5-5","source":"probe"},"effort":{"value":"xhigh","source":"environment"},"role":{"value":"implementer","source":"inferred","confidence":"medium","rule_version":"v2","evidence":["claim:pm-gh1642","claim:pm-gh1643","claim:pm-gh1644","claim:pm-hke9z7","lineage:pm-hke9z7","lineage:pm-5k4v","lineage:pm-doxj"]},"topic":{"value":"workset:pm-gh1642+pm-gh1643+pm-gh1644+pm-hke9z7","source":"inferred","confidence":"medium","rule_version":"v2","evidence":["claim:pm-gh1642","claim:pm-gh1643","claim:pm-gh1644","claim:pm-hke9z7","lineage:pm-hke9z7","lineage:pm-5k4v","lineage:pm-doxj"]},"version":{"value":"2.1.314","source":"probe"}},"op":"comment_add","patch":[{"op":"add","path":"/metadata/comments/258","value":{"created_at":"2026-10-31T22:12:47.451Z","author":"harness:claude-code","text":"October 31 checkpoint (verified live 22:10 UTC): the 03:45 UTC manual run 41165698315 and the 08:47 UTC scheduled run 41185994587 both stopped at static-quality-gate on 1a72d089d, before publication. The causes were two registry advisories published that morning (follow-redirects GHSA-mrwp-f3q8-3g9p, minimatch GHSA-g3hv-75wg-m249) and the pm-changelog 2026.10.31 grammar census drift. PR #1638 (3ff44fb6a) fixed both. Closing blocker #1639 at 19:16 UTC fired Auto Release (blocker_retry) 41233462967 → Release 41236628626 → v2026.10.31: npm latest 2026.10.31, GitHub Release published 20:51 UTC. Today has exactly one cut and no duplicate automatic release, and the nine other blocker_retry runs were skipped. The nightly quality alert #1641 had the same root cause and is closed."}},{"op":"replace","path":"/metadata/updated_at","value":"2026-10-31T22:12:47.452Z"}],"before_hash":"726a9d13392bbff0d67d3ef6335573bd5af0e5d687a1d5d5ced38a40cb0a0a33","after_hash":"6d04e973eae1c0d9d96628dae8004240edaa87f87417fc5635e5d06a303f4517","item_hash_version":3,"context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"},"version":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"61afdba7c0cf5bad087c45a9ae165d91962a9db958c0b42e9260221a51c81961"}
+{"hash_algorithm":"sha256","ts":"2026-11-01T01:17:28.301Z","author":"harness:claude-code","author_source":"detected","agent_harness":"claude-code","agent_model":"claude-opus-5-5","agent_model_source":"probe","agent_instance":"8a4268fe12c2a5f9fcd4da8d","agent_provenance":{"model":{"value":"claude-opus-5-5","source":"probe"},"effort":{"value":"xhigh","source":"environment"},"role":{"value":"implementer","source":"argv"},"topic":null,"version":{"value":"2.1.314","source":"probe"}},"op":"update","patch":[{"op":"replace","path":"/metadata/expected_result","value":"At November 2 06:00 UTC verify the November 1 daily cut against the latest merged delivery, including tag/npm identity, npm/Bun/npx/bunx consumers and fresh observability; preserve original reliability outcomes."},{"op":"replace","path":"/metadata/reminders/0/text","value":"Verify the 2026-11-01 daily cut (the first after PR #1645) against npm latest, GitHub Release, npm/npx/Bun/bunx consumers and fresh Sentry/telemetry; record any scheduled failure stage"},{"op":"replace","path":"/metadata/reminders/0/at","value":"2026-11-02T06:00:00.000Z"},{"op":"replace","path":"/metadata/updated_at","value":"2026-11-01T01:17:28.301Z"}],"before_hash":"6d04e973eae1c0d9d96628dae8004240edaa87f87417fc5635e5d06a303f4517","after_hash":"c407e03337718ac338b979547e99cef8482dda365ce59e1605c33d7227469f0b","item_hash_version":3,"context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"},"version":{"status":"resolved","resolver":"claude_session_file","rule_version":"v1"}}},"event_class":"maintenance","record_hash_version":1,"record_hash":"3db186f945e46857ab10b909dbec5b17abca9020bb7d34777259c6cbc5324930"}