A ref that cannot link is plain text — one ref contract for the aggregator (design)

Design doc: files/designs/code-aggregator-ref-contract.md in the keep-the-four-repos-healthy goal workspace. Filed by the design architect (pass 55, 2026-11-05). Audit HEAD: origin/main 955ed23. No user/review — the validity rule is already landed (#4952) and the layer rulings below are the architect's recorded decisions.

The defect

The aggregator's pages are stitched together with GitHub refs (owner/repo#N). A ref is typed RefKey — a promise that parseRef accepts it — but producers and gates make that promise without checking it, and renderers trust it: parseRef throws, and one bad ref takes down a whole page (/today #6174, /search + project pages #6211, pool project pages + briefs #6224), a whole board (#6210), or silently deletes the board's entire in-flight lane (#6195 — the throw is swallowed by an unlogged catch). Six filings in two days, one per producer/consumer pair, each leaving the same question open — which layer owns it? — and answering it differently: drop at a named boundary (#6174), "a design call for one deliberate pass" (#6195), both layers (#6211), gate-drops-the-item preferred (#6224). This issue is that deliberate pass: it records the disposition contract once and adopts the five open origins as its phases. Per-site fixes that choose different layers will diverge; this design is the layer ruling they share.

Evidence

The validity rule exists. pm-stack/aggregator/src/model.ts:454-467 — parseRef throws on non-owner/repo#N shapes and unsafe-integer numbers; the comment states the pm-stack(aggregator): validateNewComment/validateNewAction accept ref keys parseRef rejects (GH_REF number unbounded) #4952 rule (boundary validators route through isValidRef, never a local regex). refsInText complies (:484; pm-stack(aggregator): refsInText emits ref keys parseRef rejects — builder-made session records fail isSessionRecord and drop whole #4890 closed today was its producer instance). Render embodiments exist: PR fix(aggregator): Needs-you render no longer 500s on a tier item with a non-GitHub ref #6149's isValidRef guard → plain text (render.ts:2489-2524), sessionRecordLine's try/catch (render.ts:1912). The overview instance pm-stack(aggregator): overview front door throws on a lined-up task with a non-GitHub ref — reconciledInProgress reaches refUrl unguarded #6205 closed in-window today.
Gates shape-check or ignore ref fields. isLinedUpElement (feed.ts:223) never checks ref; assembleWorkItems copies it verbatim (work-items.ts:191). isProjectPoolItem (projects.ts:280-283) checks parentRef/epicRef as any string, blockedBy/blocks as any string array.
Producers mint/carry unvalidated refs. titleRefs (sources/live.ts:538) adds its explicit-ref and alias-shorthand candidates unguarded (pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210). projectorLinks (sources/live.ts:484) maps registry rows verbatim (pm-stack(aggregator): a projector registry link with an unsafe-integer ref 500s /search and the project page — projector gate admits it, projectorLinks never validates, renderSearch + roadmap rows call refUrl unguarded #6211), and the projector's POST /links gate is a local regex whose \d+ admits unsafe numbers (pm-stack/projector/src/server.ts:246) — the anti-pattern the pm-stack(aggregator): validateNewComment/validateNewAction accept ref keys parseRef rejects (GH_REF number unbounded) #4952 comment names, in the sibling app; legacy pmp_links rows bypass even that.
Consumers throw. 25 refUrl( sites in pm-stack/aggregator/src; the filed unguarded ones on stored/feed-derived refs are at render.ts:3368 (pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174), render.ts:1506 + assembly parseRef at sources/live.ts:634 (pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195/pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210), render.ts:510 + :6741 (pm-stack(aggregator): a projector registry link with an unsafe-integer ref 500s /search and the project page — projector gate admits it, projectorLinks never validates, renderSearch + roadmap rows call refUrl unguarded #6211), brief.ts:96 + render.ts:6479/:6590/:6645/:6648 (pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224). The board lane's catch (sources/live.ts:1177-1184) swallows the throw with no log line.

The contract (recorded decisions)

A ref is an annotation, not identity; validity is isValidRef; an invalid ref is dropped where it is consumed and rendered as plain text where it is displayed — it never throws, and it never takes its item with it.

One validity rule — isValidRef only, in both apps; no local regex re-derives it. parseRef keeps throwing for internally constructed refs.
Drop the ref, keep the item (stored/feed data) — at the consumption step beside each gate (normalizeOrchestrationSection, poolFromFeed, projectorLinks), not inside the pure predicates: an invalid ref field/edge is removed, the element survives, the drop is counted/logged. This pins against pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224's drop-the-whole-item gate option (an edge is an element of the item; design(aggregator): ingest, fail-closed — one validation contract for every document the PM stack consumes #2576's drop-element-and-count applies at that granularity — item loss over an annotation is the pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195 harm). Direct user POSTs whose ref is the subject keep pm-stack(aggregator): validateNewComment/validateNewAction accept ref keys parseRef rejects (GH_REF number unbounded) #4952's landed reject behaviour.
Producers mint only valid refs — every scrape/derivation routes candidates through isValidRef; an unfiltered ref scrape is a defect per se.
Rendering and deriving never throw on a ref — stored/feed-derived refs reach refUrl only through one safe-link helper (valid → anchor, invalid → plain text, the fix(aggregator): Needs-you render no longer 500s on a tier item with a non-GitHub ref #6149 treatment); derive paths treat an invalid edge as absent. Defence in depth behind rule 2, not a substitute.
Drops are observable — every rule-2/3 drop leaves a log line or provenance/count note; assembly catches that can hide a lane log what they swallowed.

Phases (each independently landable; origins adopted, no parallels)

Phase 1 — the lined-up seam: pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174 + pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195 (consumption-step drop + objectives/board consumption via the safe treatment + lane catch logging). pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174's "chosen layer" question is answered by the contract: consumption step + render helper, not the predicate, not render alone.
Phase 2 — minting and the registry: pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210 + pm-stack(aggregator): a projector registry link with an unsafe-integer ref 500s /search and the project page — projector gate admits it, projectorLinks never validates, renderSearch + roadmap rows call refUrl unguarded #6211 (titleRefs filtering; projectorLinks consumption filtering incl. legacy rows; projector POST /links gate aligned; /search + roadmap rendering via the helper).
Phase 3 — pool edges: pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224 (edges drop / single-ref fields clear at poolFromFeed, item retained per rule 2; brief + graph consumers per rule 4). Its dispatch implementation copy Fix #6224: pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge #6226 is this phase's vehicle if its holder takes the pin; otherwise triage reconciles the two so the fix lands once.
Phase 4 — closing inventory (filed with this design, gated on Phases 1–3): every refUrl/parseRef consumer and ref-carrying gate/scrape/store path in aggregator + projector dispositioned; contract statement in model.ts beside the pm-stack(aggregator): validateNewComment/validateNewAction accept ref keys parseRef rejects (GH_REF number unbounded) #4952 comment; named dispositions for queue-view.ts, sources/live.ts:368, comments.ts:494, render.ts:3574 + :6756, the sessionRecordLine alignment, and a stored-pmp_links recount.

Open questions stay in the design doc (a typed non-GitHub ref kind for goal ids; drop-count surfacing — shared with #2576's dashboard question; projector bulk-import paths).

Acceptance

All four phases closed or dispositioned with evidence.
No stored/feed-derived ref can throw a render or a derivation anywhere in the aggregator (Phase 4 inventory is the proof).
Every invalid-ref drop in the filed surfaces leaves the trace rule 5 requires.

Acceptance criteria

One ref contract is documented for the aggregator: when a ref can link, when it must render as plain text, and why.
Every surface that renders refs (board, in-flight, detail views) follows the same contract, with tests for a linkable ref and an unlinkable ref.
No ref renders as a dead/broken link anywhere the contract applies.

CODE-1825

🤖 Agent note · design filed

Agent: design architect (repo-design-architect-code), pass 55
Found via: audit of the aggregator ref family against origin/main 955ed23 — six filings in two days of one defect (pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174, pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195, pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210, pm-stack(aggregator): a projector registry link with an unsafe-integer ref 500s /search and the project page — projector gate admits it, projectorLinks never validates, renderSearch + roadmap rows call refUrl unguarded #6211, pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224 open; pm-stack(aggregator): overview front door throws on a lined-up task with a non-GitHub ref — reconciledInProgress reaches refUrl unguarded #6205 closed in-window today), each leaving the layer question open and answering it differently in its acceptance.
Task: this issue is the design of record for the aggregator ref contract. Phases: Phase 1 = pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174 + pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195 (adopted), Phase 2 = pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210 + pm-stack(aggregator): a projector registry link with an unsafe-integer ref 500s /search and the project page — projector gate admits it, projectorLinks never validates, renderSearch + roadmap rows call refUrl unguarded #6211 (adopted), Phase 3 = pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224 (adopted; dispatch copy Fix #6224: pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge #6226 is its vehicle if its holder takes the pin), Phase 4 = Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 (closing inventory, NEW, gated on Phases 1–3). All six verified in this issue's sub_issues list (count=6).
Recorded decisions: the five contract rules in this issue's body — one validity rule (isValidRef); drop the ref at the consumption step, keep the item (pinning against pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224's drop-the-item option); producers mint only valid refs; rendering/deriving via one safe-link helper, never a throw; drops observable.
Design doc: files/designs/code-aggregator-ref-contract.md (keep-the-four-repos-healthy goal workspace); register entry added in files/potential-projects.md.
Expectation: triage of the phases records the phase structure instead of re-filing or re-scoping; Phase 4 does not start before Phases 1–3 land.

🤖 Agent note · Phase 4 closing state (from task 13017, lane code-g)

Phase 4 of this design (issue #6243) is complete except for merge: the closing inventory is posted on #6243 (every refUrl/parseRef consumer and ref-carrying gate/scrape/store path in pm-stack/aggregator + pm-stack/projector dispositioned, none unclassified; all seven required dispositions recorded; pmp_links recount 0 rows / 0 invalid), the contract statement is landed in pm-stack/aggregator/src/model.ts beside the #4952 comment, and disposition 5 (sessionRecordLine) is aligned to the rule-4 plain-text treatment — all in PR #7642 with pinning tests (18 pass). No follow-up issues were owed by the inventory. When PR #7642 merges, all four phases of the ref contract are closed out.

🤖 Agent note · picked up

Agent: queue worker (repo-worker-code), lane code-t
Session / ID: task 12925 (resumed)
Found via: queue claim
Task: task 12925 remainder for issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 (Phase 4 inventory fixes for issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243, branch queue/12925-6242-phase4-inventory) was left at awaiting_ci by the previous holder; resuming from that checkpoint — verify checks, merge PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639, close issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 and issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242, complete task 12925. (This note also covers the parked agent note the previous holder recorded as owed, OB-2026-11-08-011, when both GitHub write buckets were rate-limited.)
Verification so far: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 open, mergeable, 3 files / +156/-5, diff matches the Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 ref-contract intent; no Codex reviews or review comments exist on PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639, so no undispositioned Codex findings; prior holder recorded ref-contract-6243 tests 5/5 passing.
Expectation: merge PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 once its basic checks are green, then close issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 and issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with evidence.
Links: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 · issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 · issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243

🤖 Agent note · parked

Agent: queue worker (repo-worker-code), lane code-t
Session / ID: task 12925
Found via: queue claim (resumed from the previous holder's awaiting_ci checkpoint)
Task: task 12925 parked on CI: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639's basic checks (scripts-test 3.12/3.14, proxy port fallbacks, fleet-search parity) have been QUEUED since 2026-11-08T10:31Z and were still queued after a 5-minute poll at 11:19Z — a runner backlog, not a failure. No check has concluded against PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639.
Verified this shift: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 open + mergeable, diff matches issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243's ref-contract intent, no Codex findings exist to disposition, prior holder's targeted tests (ref-contract-6243, 5/5) recorded in the task checkpoint.
Expectation / resume condition: when the basic checks on PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 conclude green, merge it, close issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 and issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with evidence, and complete task 12925.
Links: PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 · issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 · issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243

🤖 Agent note · parked

Agent: queue worker (repo-worker-code), lane code-u
Session / ID: task 12925
Found via: queue claim
Task: Parent completion vehicle for design epic issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 (aggregator ref contract): Phase 2 via PR fix(aggregator): filter invalid title refs #6475 / issue pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210, Phase 4 via issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243, then close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242.
Expectation: Fresh state 2026-11-08T11:44Z: PR fix(aggregator): filter invalid title refs #6475 is MERGED and issue pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210 is CLOSED — Phase 2 is complete, acceptance item 1 satisfied by landed work. Phase 4 (issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243) is actively owned by task 13017: its closing inventory is posted on issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 and its implementation sits in PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642, parked awaiting CI (all basic checks QUEUED in the fleet runner backlog at this check; no Codex findings posted). Lane code-u does not duplicate task 13017's work. Resume condition: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 merges and issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 closes — then close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with the phase evidence and complete task 12925.
Links: issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242, issue pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210, PR fix(aggregator): filter invalid title refs #6475, issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243, PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642

🤖 Agent note · parked

Agent: repo queue worker, lane code-n (repo-worker-code)
Session / ID: cron repo-queue-worker-shift-n run 2026-11-08 05:27 CST; task 12925
Found via: claimed task 12925 this shift; execution-lane assignment returned direct (row 4635), so this lane resumed it from the prior verified checkpoint
Task: task 12925 for issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 (small_fix remainder). Fresh revalidation: Phase 2 is done — PR fix(aggregator): filter invalid title refs #6475 merged and issue pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210 is CLOSED. The remaining chain is unchanged and still gated: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 (Phase 4 inventory for issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243, owned by task 13017) is OPEN, MERGEABLE, with all basic checks QUEUED (fleet CI backlog); issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 is OPEN. Nothing in this task can advance until that PR lands, and it belongs to another live task — no parallel work started.
Expectation: resume when PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 merges and issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 closes; then close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with the four-phase evidence. Parked 1200s.
Links: task 12925 · issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 · issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 · PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 · task 13017

🤖 Agent note · parked

Agent: queue worker (repo-worker-code), lane code-x
Session / ID: task 12925
Found via: queue claim · resume
Task: Close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 once its Phase 4 lands — resume condition re-checked fresh at 2026-11-08T12:21Z: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 still OPEN (checks QUEUED, owned by task 13017, parked awaiting CI), issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 still OPEN. Gate not yet met, nothing to close yet.
Expectation: Resume when PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 merges and issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 closes; then close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with the Phase 1–4 evidence.
Links: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 · issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 · issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242

🤖 Agent note · parked

Agent: repo-worker-code (lane V, queue task 12925, small_fix)
Session / ID: cron repo-queue-worker-shift-v, run 2026-11-08 ~06:40 CST
Found via: queue claim · resume
Task: Close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 once its Phase 4 lands — resume condition re-checked fresh at 2026-11-08T12:40Z: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 still OPEN (owned by task 13017), issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 still OPEN. Gate not yet met; task 12925 parked 1200s naming the same condition. No parallel work started.
Expectation: Resume when PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 merges and issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 closes; then close issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242 with the Phase 1–4 evidence.
Links: PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 · issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 · issue A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242

🤖 Agent note · done — all four phases closed, design contract landed

Agent: queue worker (repo-worker-code), lane code-u, task 12925
Phase evidence (verified fresh 2026-11-08): Phase 1 + Phase 3 closed earlier (issues pm-stack(aggregator): /today render 500s on a lined-up task with a non-GitHub ref — objectives call refUrl unguarded and the section gate never checks linedUp ref #6174, pm-stack(aggregator): one lined-up task with a non-GitHub ref silently deletes the board's whole in-flight lane — inFlightBoardItems parseRef throws into an unlogged catch #6195, pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge 500s the project page and throws the brief — isProjectPoolItem validates edges by shape only, brief + pool-graph + epic-graph call refUrl unguarded #6224, Fix #6224: pm-stack(aggregator): a pool item with an invalid blockedBy/blocks edge #6226 CLOSED). Phase 2: PR fix(aggregator): filter invalid title refs #6475 merged and issue pm-stack(aggregator): board render throws on a lined-up task whose title carries an unsafe-integer ref in a non-first link — titleRefs mints refs parseRef rejects, inFlightBoardItems validates only the first #6210 CLOSED. Phase 4: issue Closing ref inventory — disposition every ref consumer and producer against the ref contract (Phase 4 of #6242) #6243 CLOSED — closing inventory landed via PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 (merged 2026-11-08, merge sha 8a0a1c0), including the disposition-5 sessionRecordLine alignment salvaged from duplicate PR feat(pm-aggregator): ref contract statement + sessionRecordLine plain-text alignment (#6243) #7642 (closed superseded; salvage note PR fix(aggregator): close the ref contract — Phase 4 inventory fixes (#6243) #7639 comment 6721827851). Tests: 60/60 across 7 aggregator files on the laptop worktree at merge.
All acceptance criteria of task 12925 are met; closing this design issue on its merits (landed work), per the phase evidence above.

Navigation Menu

Repository navigation

A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242

The defect

Evidence

The contract (recorded decisions)

Phases (each independently landable; origins adopted, no parallels)

Acceptance

Acceptance criteria

38 remaining items

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

A ref that cannot link is plain text — one ref contract for the aggregator (design) #6242

Description

The defect

Evidence

The contract (recorded decisions)

Phases (each independently landable; origins adopted, no parallels)

Acceptance

Acceptance criteria

Activity

developing-today commented on Nov 5, 2026

dezren39 commented on Nov 5, 2026

38 remaining items

developing-today commented on Nov 8, 2026

dezren39 commented on Nov 8, 2026

dezren39 commented on Nov 8, 2026

dezren39 commented on Nov 8, 2026

developing-today commented on Nov 8, 2026

dezren39 commented on Nov 8, 2026

dezren39 commented on Nov 8, 2026

developing-today commented on Nov 8, 2026

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions