Refuse malformed settings writes with typed recovery

.agents/pm/extensions/.managed-extensions.json

@@ -1,6 +1,6 @@
 {
   "version": 1,
-  "updated_at": "2026-09-27T10:35:46.561Z",
+  "updated_at": "2026-09-27T11:48:50.897Z",
   "entries": [
     {
       "name": "pm-changelog",

.agents/pm/history/pm-aao1hy.jsonl

@@ -0,0 +1 @@
+{"hash_algorithm":"sha256","ts":"2026-09-27T12:25:46.741Z","author":"harness:codex","author_source":"detected","agent_harness":"codex","agent_model":"gpt-6-sol","agent_model_source":"probe","agent_instance":"d47f0e0d79925f50a0df2359","agent_provenance":{"model":{"value":"gpt-6-sol","source":"probe"},"effort":{"value":"high","source":"probe"},"role":{"value":"implementer","source":"argv"},"topic":null},"op":"create","patch":[{"op":"add","path":"/metadata/id","value":"pm-aao1hy"},{"op":"add","path":"/metadata/title","value":"GH-1328: linked-test execution does not disclose item evidence recording"},{"op":"add","path":"/metadata/description","value":"GitHub issue #1328 reports that pm test <id> --run says changed:false both when default test-result-tracking disables item recording and when enabled tracking appends test_run_track. Background runs also report passed without item evidence. Source: https://github.com/unbraind/pm-cli/issues/1328"},{"op":"add","path":"/metadata/type","value":"Issue"},{"op":"add","path":"/metadata/status","value":"open"},{"op":"add","path":"/metadata/priority","value":1},{"op":"add","path":"/metadata/tags","value":["agent-ux","area:cli","area:testing","defect","github"]},{"op":"add","path":"/metadata/created_at","value":"2026-09-27T12:25:46.741Z"},{"op":"add","path":"/metadata/updated_at","value":"2026-09-27T12:25:46.741Z"},{"op":"add","path":"/metadata/author","value":"harness:codex"},{"op":"add","path":"/metadata/acceptance_criteria","value":"Reproduce disabled and enabled policy paths in disposable trackers; add an explicit recorded state and reason or accurate changed field; verify foreground and background item history plus SDK and CLI parity; pass exact coverage"},{"op":"add","path":"/metadata/goal","value":"project management = context management"},{"op":"add","path":"/metadata/objective","value":"Linked-test results explicitly say whether execution evidence reached the item history"},{"op":"add","path":"/metadata/value","value":"Agents can trust closeout evidence without inspecting a separate history stream"},{"op":"add","path":"/metadata/impact","value":"Prevent false assumptions that a passing linked test was recorded"},{"op":"add","path":"/metadata/outcome","value":"Foreground and background test runs expose an accurate recording receipt and recovery command"},{"op":"add","path":"/metadata/parent","value":"pm-lm0j"},{"op":"add","path":"/metadata/risk","value":"high"},{"op":"add","path":"/metadata/severity","value":"high"},{"op":"add","path":"/metadata/repro_steps","value":"Run pm test <id> --run with default tracking and then enabled tracking; both currently print changed:false although only the second appends test_run_track."},{"op":"add","path":"/metadata/expected_result","value":"The run result accurately reports recording and the command to enable it when disabled."},{"op":"add","path":"/metadata/actual_result","value":"changed:false is printed in both cases without recording provenance."},{"op":"add","path":"/metadata/affected_version","value":"2026.9.27"},{"op":"add","path":"/metadata/component","value":"linked-test runtime and item evidence"},{"op":"add","path":"/metadata/dependencies","value":[{"id":"pm-16f4","kind":"discovered_from","created_at":"2026-09-27T12:25:46.741Z","author":"harness:codex","source_kind":"cli:create:dep","author_source":"detected"}]},{"op":"add","path":"/metadata/comments","value":[{"created_at":"2026-09-27T12:25:46.741Z","author":"harness:codex","text":"Duplicate check: all-status GH-1328 and test-result-tracking searches found the closed tracking feature pm-16f4 and its parent epic but no open owner for the false recording receipt."}]}],"before_hash":"3cc22dff72be7b14824654a7a64ea62b04799939b2fee54c1b5f52ca60bf6df0","after_hash":"165d3d0e41e03ba21fbf99aa28cec9057a0e8b98832ab39d3533484d38867451","item_hash_version":3,"message":"","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"effort":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"version":{"status":"unavailable","reason":"harness_unavailable","resolver":"ai_agent_version","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"2cbeedc45af5f49f8562a96c6e41fcfdf6dae3b5edf035a68e6e7a91ce7112ea"}

.agents/pm/history/pm-c3aiik.jsonl

@@ -0,0 +1 @@
+{"hash_algorithm":"sha256","ts":"2026-09-27T12:26:03.741Z","author":"harness:codex","author_source":"detected","agent_harness":"codex","agent_model":"gpt-6-sol","agent_model_source":"probe","agent_instance":"d47f0e0d79925f50a0df2359","agent_provenance":{"model":{"value":"gpt-6-sol","source":"probe"},"effort":{"value":"high","source":"probe"},"role":{"value":"implementer","source":"argv"},"topic":null},"op":"create","patch":[{"op":"add","path":"/metadata/id","value":"pm-c3aiik"},{"op":"add","path":"/metadata/title","value":"GH-1325: history diff cursor drops newest rows and strands the final page"},{"op":"add","path":"/metadata/description","value":"GitHub issue #1325 reproduces default-budget pm history <id> --diff --format json with a cursor that advances compact_history but not diff. Later pages repeat older diffs and can report has_more:true without next_cursor; close and release diffs become unreachable. Source: https://github.com/unbraind/pm-cli/issues/1325"},{"op":"add","path":"/metadata/type","value":"Issue"},{"op":"add","path":"/metadata/status","value":"open"},{"op":"add","path":"/metadata/priority","value":1},{"op":"add","path":"/metadata/tags","value":["agent-ux","area:history","area:output","defect","github"]},{"op":"add","path":"/metadata/created_at","value":"2026-09-27T12:26:03.741Z"},{"op":"add","path":"/metadata/updated_at","value":"2026-09-27T12:26:03.741Z"},{"op":"add","path":"/metadata/author","value":"harness:codex"},{"op":"add","path":"/metadata/acceptance_criteria","value":"Reproduce with disposable 20-entry history and real tracker example; advance or independently continue all truncated diff rows; never emit has_more without a cursor; verify all rows and snapshot contracts with exact coverage"},{"op":"add","path":"/metadata/goal","value":"project management = context management"},{"op":"add","path":"/metadata/objective","value":"Every bounded history diff can be fully reconstructed through continuation cursors"},{"op":"add","path":"/metadata/value","value":"Agents can recover the actual latest lifecycle and audit changes"},{"op":"add","path":"/metadata/impact","value":"Prevent false conclusions that an item was never closed or released"},{"op":"add","path":"/metadata/outcome","value":"Diff and compact history advance consistently and final pages have truthful continuation receipts"},{"op":"add","path":"/metadata/parent","value":"pm-5t33or"},{"op":"add","path":"/metadata/risk","value":"high"},{"op":"add","path":"/metadata/severity","value":"high"},{"op":"add","path":"/metadata/repro_steps","value":"Page pm history <id> --diff --format json with default budget using --output-cursor; diff repeats earlier indices and the final page has has_more without next_cursor."},{"op":"add","path":"/metadata/expected_result","value":"Every diff row is reachable exactly once and the final page has no dangling has_more."},{"op":"add","path":"/metadata/actual_result","value":"Newest diff rows are unreachable without unbounded output."},{"op":"add","path":"/metadata/affected_version","value":"2026.9.27"},{"op":"add","path":"/metadata/component","value":"history output budgets and continuation"},{"op":"add","path":"/metadata/dependencies","value":[{"id":"pm-tqel","kind":"discovered_from","created_at":"2026-09-27T12:26:03.741Z","author":"harness:codex","source_kind":"cli:create:dep","author_source":"detected"}]},{"op":"add","path":"/metadata/comments","value":[{"created_at":"2026-09-27T12:26:03.741Z","author":"harness:codex","text":"Duplicate check: all-status GH-1325 and cursor-diff searches found the general output-budget feature pm-5t33or and closed history grammar work pm-tqel; neither owns this specific paired-diff continuation failure."}]}],"before_hash":"3cc22dff72be7b14824654a7a64ea62b04799939b2fee54c1b5f52ca60bf6df0","after_hash":"13bc8c72eab5e788e6cc40fb9749d2d0592eee9c03f298019a5000def3ef0afe","item_hash_version":3,"message":"","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"effort":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"version":{"status":"unavailable","reason":"harness_unavailable","resolver":"ai_agent_version","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"8a54c7ecfd9f472b7a3ccf66368b4825d95e17d11a0aaab547da9c4846f4029d"}

.agents/pm/history/pm-ft90q2.jsonl

@@ -0,0 +1 @@
+{"hash_algorithm":"sha256","ts":"2026-09-27T12:26:45.447Z","author":"harness:codex","author_source":"detected","agent_harness":"codex","agent_model":"gpt-6-sol","agent_model_source":"probe","agent_instance":"d47f0e0d79925f50a0df2359","agent_provenance":{"model":{"value":"gpt-6-sol","source":"probe"},"effort":{"value":"high","source":"probe"},"role":{"value":"implementer","source":"argv"},"topic":null},"op":"create","patch":[{"op":"add","path":"/metadata/id","value":"pm-ft90q2"},{"op":"add","path":"/metadata/title","value":"GH-1327: Windows nightly static inventory unreadable-source assertion fails"},{"op":"add","path":"/metadata/description","value":"GitHub issue #1327 records Nightly Validation run 36310799887 at main 2b14d2fdee9d65dd1699d79e617ab53a0ae91dc7. Windows Node 24 shard 2/2 failed the read-only static extension inventory unreadable-roots/sources integration case: expected managed_state_unreadable plus source detail but received extensions_unreadable. This is separate from shard 1 init test timeouts. Source: https://github.com/unbraind/pm-cli/issues/1327"},{"op":"add","path":"/metadata/type","value":"Issue"},{"op":"add","path":"/metadata/status","value":"open"},{"op":"add","path":"/metadata/priority","value":1},{"op":"add","path":"/metadata/tags","value":["area:extensions","ci","defect","github","nightly","node24","windows"]},{"op":"add","path":"/metadata/created_at","value":"2026-09-27T12:26:45.447Z"},{"op":"add","path":"/metadata/updated_at","value":"2026-09-27T12:26:45.447Z"},{"op":"add","path":"/metadata/author","value":"harness:codex"},{"op":"add","path":"/metadata/acceptance_criteria","value":"Inspect run 36310799887 shard 2 and reproduce on hosted Windows Node 24; distinguish platform error mapping from fixture assumptions; fix SDK or test according to real behavior; verify no activation or writes and rerun nightly plus exact coverage"},{"op":"add","path":"/metadata/goal","value":"project management = context management"},{"op":"add","path":"/metadata/objective","value":"Static extension inventory reports precise unreadable-source receipts consistently on Windows"},{"op":"add","path":"/metadata/value","value":"Preserve trustworthy read-only extension diagnostics across platforms"},{"op":"add","path":"/metadata/impact","value":"Avoid masking a missing managed-state reason behind a generic extensions_unreadable result"},{"op":"add","path":"/metadata/outcome","value":"Windows nightly shard 2 verifies precise refusal semantics without altering inventory safety"},{"op":"add","path":"/metadata/parent","value":"pm-ul9rye"},{"op":"add","path":"/metadata/risk","value":"high"},{"op":"add","path":"/metadata/severity","value":"high"},{"op":"add","path":"/metadata/repro_steps","value":"Inspect Nightly run 36310799887 shard 2/2: static-extension-inventory.integration.spec.ts expects managed_state_unreadable and source detail but receives extensions_unreadable."},{"op":"add","path":"/metadata/expected_result","value":"Precise managed-state and source unreadable receipts match the tested read-only conditions."},{"op":"add","path":"/metadata/actual_result","value":"Generic extensions_unreadable replaces expected receipts on Windows; shard 2 fails."},{"op":"add","path":"/metadata/affected_version","value":"2026.9.27"},{"op":"add","path":"/metadata/component","value":"static extension inventory Windows nightly"},{"op":"add","path":"/metadata/dependencies","value":[{"id":"pm-lhhnx9","kind":"discovered_from","created_at":"2026-09-27T12:26:45.447Z","author":"harness:codex","source_kind":"cli:create:dep","author_source":"detected"}]},{"op":"add","path":"/metadata/comments","value":[{"created_at":"2026-09-27T12:26:45.447Z","author":"harness:codex","text":"Duplicate check: GH-1327 and all-status static inventory and Windows nightly searches found closed pm-lhhnx9 and older nightly lineages but no owner for this exact Windows assertion. Run 36310799887 log proves a distinct shard-2 failure."}]}],"before_hash":"3cc22dff72be7b14824654a7a64ea62b04799939b2fee54c1b5f52ca60bf6df0","after_hash":"7a0605f579e3a9ad3d6e76841c8143446ea67db7143b70ed6587c992342fb147","item_hash_version":3,"message":"","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"effort":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"version":{"status":"unavailable","reason":"harness_unavailable","resolver":"ai_agent_version","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"e78823eed01e9a800143e3142a99e6bd2f45c16a00cf94704de6fd06e9c8757c"}

.agents/pm/history/pm-kvhnb5.jsonl

@@ -0,0 +1,2 @@
+{"hash_algorithm":"sha256","ts":"2026-09-27T12:26:27.673Z","author":"harness:codex","author_source":"detected","agent_harness":"codex","agent_model":"gpt-6-sol","agent_model_source":"probe","agent_instance":"d47f0e0d79925f50a0df2359","agent_provenance":{"model":{"value":"gpt-6-sol","source":"probe"},"effort":{"value":"high","source":"probe"},"role":{"value":"implementer","source":"argv"},"topic":null},"op":"create","patch":[{"op":"add","path":"/metadata/id","value":"pm-kvhnb5"},{"op":"add","path":"/metadata/title","value":"GH-1326: Windows nightly bundled-package init test exceeds 30 seconds"},{"op":"add","path":"/metadata/description","value":"GitHub issue #1326 records Nightly Validation run 36310799887 at main 2b14d2fdee9d65dd1699d79e617ab53a0ae91dc7. Windows Node 24 shard 1/2 failed two assertions in tests/unit/commands/workspace/init-command.spec.ts because the bundled first-party package initialization test timed out at 30000 ms. This is separate from shard 2 static inventory assertion. Source: https://github.com/unbraind/pm-cli/issues/1326"},{"op":"add","path":"/metadata/type","value":"Issue"},{"op":"add","path":"/metadata/status","value":"open"},{"op":"add","path":"/metadata/priority","value":1},{"op":"add","path":"/metadata/tags","value":["area:packages","ci","defect","github","nightly","node24","windows"]},{"op":"add","path":"/metadata/created_at","value":"2026-09-27T12:26:27.673Z"},{"op":"add","path":"/metadata/updated_at","value":"2026-09-27T12:26:27.673Z"},{"op":"add","path":"/metadata/author","value":"harness:codex"},{"op":"add","path":"/metadata/acceptance_criteria","value":"Inspect the two timeout cases from run 36310799887; reproduce on hosted Windows Node 24; fix the underlying package or test budget with negative control; rerun full nightly matrix and exact coverage"},{"op":"add","path":"/metadata/goal","value":"project management = context management"},{"op":"add","path":"/metadata/objective","value":"Nightly Windows package initialization tests distinguish product failures from legitimate runtime budget needs"},{"op":"add","path":"/metadata/value","value":"Keep release validation reliable without masking package install failures"},{"op":"add","path":"/metadata/impact","value":"Avoid an untrusted nightly release signal after timed-out bundled package initialization"},{"op":"add","path":"/metadata/outcome","value":"The exact Windows nightly test completes or reports a specific actionable package failure"},{"op":"add","path":"/metadata/parent","value":"pm-ul9rye"},{"op":"add","path":"/metadata/risk","value":"high"},{"op":"add","path":"/metadata/severity","value":"high"},{"op":"add","path":"/metadata/repro_steps","value":"Inspect Nightly run 36310799887 shard 1/2: both bundled first-party init assertions time out at 30000 ms in init-command.spec.ts."},{"op":"add","path":"/metadata/expected_result","value":"Bundled package init completes and assertions verify installed packages on Windows."},{"op":"add","path":"/metadata/actual_result","value":"Two test cases time out after 30 seconds; Windows nightly shard 1/2 fails."},{"op":"add","path":"/metadata/affected_version","value":"2026.9.27"},{"op":"add","path":"/metadata/component","value":"Windows nightly bundled package initialization"},{"op":"add","path":"/metadata/dependencies","value":[{"id":"pm-z3ez","kind":"discovered_from","created_at":"2026-09-27T12:26:27.673Z","author":"harness:codex","source_kind":"cli:create:dep","author_source":"detected"}]},{"op":"add","path":"/metadata/comments","value":[{"created_at":"2026-09-27T12:26:27.673Z","author":"harness:codex","text":"Duplicate check: GH-1326 and all-status Windows nightly searches found historical closed failures including pm-gh1075 and pm-ul9rye; this exact run and bundled init timeout have no owner. Run 36310799887 log establishes a distinct shard-1 failure."}]}],"before_hash":"3cc22dff72be7b14824654a7a64ea62b04799939b2fee54c1b5f52ca60bf6df0","after_hash":"3e319abcbfeb3ac46b482dba4983838823fb4c312ddd5b9864c8742a9372f47a","item_hash_version":3,"message":"","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"effort":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"version":{"status":"unavailable","reason":"harness_unavailable","resolver":"ai_agent_version","rule_version":"v1"}}},"event_class":"substantive","record_hash_version":1,"record_hash":"6b83a0d68f3dd14319795d472d3698ea728a512f859491b4175559fee1fea7c4"}
+{"hash_algorithm":"sha256","ts":"2026-09-27T12:27:49.590Z","author":"harness:codex","author_source":"detected","agent_harness":"codex","agent_model":"gpt-6-sol","agent_model_source":"probe","agent_instance":"d47f0e0d79925f50a0df2359","agent_provenance":{"model":{"value":"gpt-6-sol","source":"probe"},"effort":{"value":"high","source":"probe"},"role":{"value":"implementer","source":"argv"},"topic":null},"op":"update","patch":[{"op":"add","path":"/metadata/dependencies/1","value":{"id":"pm-gh1075","kind":"discovered_from","created_at":"2026-09-27T12:27:49.229Z","author":"harness:codex","source_kind":"cli:update:dep","author_source":"detected"}},{"op":"replace","path":"/metadata/updated_at","value":"2026-09-27T12:27:49.590Z"}],"before_hash":"3e319abcbfeb3ac46b482dba4983838823fb4c312ddd5b9864c8742a9372f47a","after_hash":"174ebd6fc0dccfc34f781b75f74f6fb676a104c6f305b3edf77360dcfe8a6f69","item_hash_version":3,"message":"Link historical Windows nightly failure cited in duplicate audit","context":{"agent_provenance_outcomes":{"model":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"effort":{"status":"resolved","resolver":"codex_session_file","rule_version":"v1"},"version":{"status":"unavailable","reason":"harness_unavailable","resolver":"ai_agent_version","rule_version":"v1"}}},"event_class":"maintenance","record_hash_version":1,"record_hash":"6405ea75eae564086711926050e13aaa6c658489b2a8f969dc5ca3aa5996d4d8"}