zenmemes

New post

Feature plan: Step 1 · Step 2 · Step 3 · Step 4

Effort study: railspike-effort / v3/feedback_work_restart; host ledger. Standing P2 enrollment.

Extended estimates: enabled

Approved Step 2: 2026-11-04, no scope added since.

Cycle estimate: 0.60

Concepts touched

  • Feedback item (new): add to the concept inventory in Stage 1.

  • Feedback triage (changed), production feedback register (retired as authority): update their references in Stages 4 and 6.

  • Agent instructions (new home): docs/agents/, Stage 6.

  • Instance import: used, unchanged.

Completion Contract

  • Entry: ./v3 import-instance zenmemes.com, then ask an agent to triage feedback since a date.

  • Outcome: proposals shown; accepted changes written as item files under docs/feedback/items/; ./v3 feedback check passes; ./v3 feedback index regenerates docs/feedback/README.md; the maintainer commits.

  • Recovery: proposals are written only after review; Git reverts a bad apply; files are written atomically; rerunning a window proposes nothing already applied.

  • External verification: replay on the maintainer's WSL checkout (Stage 7).

  • Release: seeded items, spec, commands and references merged to main.

Key Risks

  • High risk: duplicate matching. Wrong attaches hide requests. Validate in Stage 5 on the October 30 window. Mitigation: uncertain matches stay separate with a review note.

  • High risk: seed loses register evidence. Validate with a two-entry pilot (Stage 3) before the rest. Mitigation: items keep current fields and link to existing history.

  • Repeated headers. The Key: value parser rejects duplicate headers, so several sources can't each be a header. Resolve in Stage 1 by putting sources in the body under a fixed section.

Stage 1

Estimate: 0.12

  • Goal: item format and a check command.

  • Dependencies: none.

  • Expected changes: spec docs/specs/feedback_item_record_v1.md (ID, Title, Kind, Workflow, Evidence, Estimate, Created headers; body with request text, Sources list, Done-when, Notes); FeedbackItemRepository::load(dir): list; FeedbackItemCheck::run(dir): list of problems; scripts/feedback.php check; ./v3 feedback dispatch; concept inventory entry.

  • Verification approach: new FeedbackItemCheckTest (valid file, malformed header, duplicate ID, source claimed twice, unknown kind or state) plus smoke set.

  • Estimate details: driver: format decisions; uncertainty: low.

  • Risks or open questions: the duplicate-header limit, resolved by the body Sources list.

  • Canonical components/API contracts touched: GenericTextRecordParser (reused); ./v3 dispatcher (extended).

Stage 2

Estimate: 0.06

  • Goal: generated index of open items.

  • Dependencies: Stage 1.

  • Expected changes: scripts/feedback.php index writes docs/feedback/README.md grouped by kind and workflow state, with a "generated; do not edit" line; atomic write.

  • Verification approach: index test (grouping, stable order, unchanged output on rerun).

  • Estimate details: driver: layout; uncertainty: low.

  • Risks or open questions: none.

  • Canonical components/API contracts touched: Stage 1 repository.

Stage 3

Estimate: 0.06

  • Goal: pilot migration of two register entries (F01 with investigation history, one bug).

  • Dependencies: Stage 1.

  • Expected changes: two item files; register sections link to them.

  • Verification approach: check passes; side-by-side review that every source, state and evidence link survives.

  • Estimate details: driver: deciding what stays in the item vs linked history; uncertainty: medium.

  • Risks or open questions:

    • Impact: lost delivery evidence.

    • Early warning / validation: the pilot review.

    • Mitigation: adjust the spec before Stage 4.

  • Canonical components/API contracts touched: item spec.

Stage 4

Estimate: 0.08

  • Goal: migrate the remaining 15 register entries; retire the register as authority.

  • Dependencies: Stage 3.

  • Expected changes: 15 item files; register gets a superseded note pointing at the index; production_feedback_management.md points at items and keeps ID and state rules.

  • Verification approach: check passes; script comparison of register IDs and source links against item files.

  • Estimate details: driver: volume; uncertainty: low.

  • Risks or open questions: none beyond Stage 3.

  • Canonical components/API contracts touched: production feedback management reference.

Stage 5

Estimate: 0.15

  • Goal: seed from the October 30 summary.

  • Dependencies: Stage 4.

  • Expected changes: proposals for all 97 catalogued rows (create or attach to an existing item); item files written after maintainer review; 10-18 and 10-21 summaries listed as unreconciled in the index.

  • Verification approach: script check that every source link in the summary appears in exactly one item; check passes; maintainer reviews the attach list.

  • Estimate details: driver: matching judgment across 97 rows; uncertainty: medium.

  • Risks or open questions:

    • Impact: wrong merges.

    • Early warning / validation: the attach list shown before writing.

    • Mitigation: separate items with a review note when unsure.

  • Canonical components/API contracts touched: item spec; check command.

Stage 6

Estimate: 0.08

  • Goal: triage procedure for new imports, in a home for agent instructions.

  • Dependencies: Stage 5.

  • Expected changes: new docs/agents/ folder for instructions agents follow when working on the project (separate from prompts/, which the application loads at runtime); docs/agents/README.md index; docs/agents/feedback_triage.md, the procedure (read imported posts in a window, match against items, present create/attach/skip proposals with skip reasons, then write and index); zenmemes_feedback_automation_checklist.md marked superseded; minimal_feedback_triage.md and production_feedback_management.md link to it; one AGENTS.md line pointing to the docs/agents/ index.

  • Verification approach: dry run on a short window already covered by Stage 5 proposes no new items; links resolve.

  • Estimate details: driver: wording the matching rules; uncertainty: low.

  • Risks or open questions: agents may still treat runtime prompts/ as the place for instructions; the README states the difference in one line.

  • Canonical components/API contracts touched: triage and management references; AGENTS.md.

Stage 7

Estimate: 0.05

  • Goal: acceptance on the maintainer's WSL checkout.

  • Dependencies: Stage 6.

  • Expected changes: none planned; fixes only if the replay fails.

  • Verification approach: maintainer runs import, triage since 2026-11-01, check and index; the two 2026-11-03 "summary field" threads become one item with two sources; a rerun proposes nothing. Evidence recorded in Step 4.

  • Estimate details: driver: real data on the real machine; uncertainty: medium (site reachability, as seen in the 2026-11-04 sweep).

  • Risks or open questions: if the import fails, record it unverified rather than substituting another environment.

  • Canonical components/API contracts touched: none new.

Enhancement Requests

  • A place for prompts and instructions (raised 2026-11-04). Disposition: docs/agents/ is created in Stage 6 with the triage procedure. Moving existing development prompts (docs/reference/ui_copy_cleanup_prompt.md, docs/plans/ui_redesign/plan_prompt.md) there is a separate small item; runtime prompts/ stays where the application expects it.

Study Snapshots

2026-11-04; scope v1; starting revision 81837cf; reference R1 not re-verified in this session. Claude, single exposed rater. Earlier snapshots: Step 1 0.55, Step 2 0.60 (preserved there).

  • Step 3 handoff: 0.60 projected total. Revised 2026-11-04 for the docs/agents/ location (Stage 6 0.07 → 0.08); cycle projection unchanged. Stage values are annotations, not summed. Not hours; no implementation or approval implied.

Status: Approved Step 3 (2026-11-04).

— Claude · 2026-11-04 UTC · Model: claude-opus-5-5