ZenUml/conf-app / .claude/skills/ai-chat-improvement
ai-chat-improvement
VerifiedBuild private cases from failed or suspicious ordinary AI Chat modifications, fix them, complete layered case validation, and run historical regression; use for continuous improvement. Use ai-chat-regression only to replay existing cases.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 5.5 KB |
| references/production-intake.md | markdown | 3.0 KB |
references/production-intake.md
View on GitHub# Ordinary AI Chat Production Data Intake
Read Diagramly `scripts/db-access.md` first. Verify that the current connection points to the intended production database, that the account or transaction is read-only, and that the UTC time window and query timeout are correct. The SQL below is a candidate filter for the current versioned-job structure; verify the live schema before running it. `Job.input`/`output` store JSON and may contain customer diagrams and complete instructions; initially return aggregate counts only.
```sql
BEGIN TRANSACTION READ ONLY;
SET LOCAL statement_timeout = '60s';
WITH scoped AS (
SELECT j.status::text AS status,
j.input->>'languageKey' AS language,
j.input->>'diagramCode' AS dsl,
j.input->>'command' AS command,
j.output->>'noChange' AS no_change
FROM "Job" AS j
WHERE j.type = 'diagram_modification'
AND j."createdAt" >= TIMESTAMPTZ '<start-utc>'
AND j."createdAt" < TIMESTAMPTZ '<end-utc>'
AND NULLIF(BTRIM(j.input->>'diagramId'), '') IS NOT NULL
AND NULLIF(BTRIM(j.input->>'diagramCode'), '') IS NOT NULL
AND NULLIF(BTRIM(j.input->>'command'), '') IS NOT NULL
AND NULLIF(BTRIM(COALESCE(j.input->>'errorMessage', '')), '') IS NULL
)
SELECT COUNT(*) FILTER (WHERE status = 'FAILED') AS failed_jobs,
COUNT(DISTINCT (language, dsl, command)) FILTER (WHERE status = 'FAILED')
AS distinct_failed_requests,
COUNT(*) FILTER (WHERE status = 'COMPLETED' AND no_change = 'true')
AS no_change_review_candidates
FROM scoped;
COMMIT;
```
These are only **versioned ordinary-modification candidates**: native Diagramly pages may call the same route. Call them conf-app traffic only after verifying a trusted external team/user mapping; `Job.source = api` is not sufficient proof. Distinguish internal tests from customer records using a verified identity allowlist; do not guess based on domain appearance. Report the counting window, timezone, query time, filters, and identity verification result. `no_change_review_candidates` is not a failure count.
After the user selects a limit, page through the required jobs into a private directory, exporting `id/createdAt/status/input/output/error` with a bounded `(createdAt,id)` cursor. Do not print raw rows to the terminal. Check export completeness and JSON parseability; record retrieval conditions, selected count, and uncovered count. If the source is only a frontend error event with no corresponding job, record the evidence gap in the flow; do not infer its DSL or instruction.
Prioritize error results with explicit user reports and `FAILED` jobs. For `COMPLETED` jobs with a created version that the user undid or reported as wrong, first verify the instruction semantics and original/output diagrams. The trigger points for `ai_chat_change_*` events may change with implementation; read the current catalog and panel code before querying. A success event reflects only its defined client-side state; it does not prove user satisfaction, acceptance, or persistence in Confluence.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack." | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |