teleo-infrastructure/docs/reports/leo-working-state-20260709/telegram-handler-direct-claim-suite-review-current.md
2026-07-13 02:32:14 +02:00

52 lines
3.4 KiB
Markdown

# Telegram Handler Direct-Claim Suite Review
Generated UTC: `2026-07-13T00:27:54Z`
## Result
- Runtime path: `pass_runtime=True`; all six direct-claim prompts returned replies from live VPS GatewayRunner using a temporary leoclean profile.
- Strict benchmark score: latest `6/6`; a prior unchanged-skill run scored `5/6`, so repeated stability is not yet proven.
- Telegram post: `false`.
- Live profile changed: `false`.
- Production DB apply executed: `false`.
- DB counts before and after: `1837,4145,4916,4670,26` -> `1837,4145,4916,4670,26`.
- Service before/after run: `active/running`, `MainPID=2403328`, `NRestarts=0`, `ExecMainStartTimestamp=Sun 2026-07-12 09:14:10 UTC`.
## Core Finding
Leo returns all six no-context replies and keeps the VPS stable. The latest
clean-session run meets the strict Cory bar, but the unchanged skill varied from
`5/6` to `6/6` across two runs. Equivalent-language scorer false negatives were
repaired locally; one stability gap remains:
- `DC-05`: the latest no-post answer includes a complete `DB readback:` receipt
and proposes a new strict staging canary. The prior no-post answer omitted the
receipt. The Telegram-visible answer has the receipt but suggests using the
strict `add_edge` path on legacy approved proposals whose payload type may not
match.
## Scenario Assessment
- `DC-01`: `pass` - Proposed/approved/applied split is correct; names two applied rows, three approved-but-unapplied rows, 14 pending, and 7 canceled; includes follow-up and authorization boundary.
- `DC-02`: `pass` - Helmer 7 Powers is approved/staged or packet-ready but not canonical; proposal `a64df080` has empty `applied_at`; canonical `public.*` rows are absent.
- `DC-03`: `pass` - Decision-matrix path is not shipped; matrix tables are absent; m3ta reviewer sign-off is not a matrix vote.
- `DC-04`: `pass` - Not a single-cause pointer diagnosis; separates Telegram/local file refs, proposal refs, document evaluations, review status, and canonical `public.sources`.
- `DC-05`: `pass_latest_unstable` - Latest no-post answer is correct; prior no-post and retained Telegram-visible variants show the contract is not yet stable.
- `DC-06`: `pass` - Direct `SOUL.md` edit is runtime/rendered artifact drift, not canonical identity; canonical change requires DB rows plus render/sync/postflight proof.
## Follow-Up Challenge
- The latest report and score supersede the older `6/6` review for current handler status.
## Next Actions
- Deploy and repeatedly no-post test the bounded `DC-05` response-contract correction.
- Reopen immediately if DC-02 stops recognizing `a64df080` as approved/unapplied, DC-03 infers matrix approval while matrix tables are absent, any DC answer omits `Next Cory-style follow-up:`, or the read-only suite mutates DB/service state.
- This pass does not authorize production apply of Helmer, mapped rich packets, Rio strategy, governance/concept schema, or any canonical `public.*` mutation.
## Evidence Files
- Direct-claim handler report: `docs/reports/leo-working-state-20260709/telegram-handler-direct-claim-suite-current.json`
- Strict score: `docs/reports/leo-working-state-20260709/telegram-handler-direct-claim-suite-score-current.md`
- Follow-up challenge report: `docs/reports/leo-working-state-20260709/telegram-handler-direct-claim-followup-current.json`
- State image: `docs/reports/leo-working-state-20260709/leo-db-state-26-direct-claim-handler.svg`