feat(consult): prompt and Skill read from the evidence card (Skill 6.9.17)

The natal system prompt now says: the card in claim_cards is the chart
evidence for this answer, quote it as given, look up one further section of
this calculation with read-consultation-evidence before writing, and read
evidence_card.backstage as a confidence cap. The "use every executed layer"
and must_use_layers sentences are replaced; local_layers paths now point at
the card. SKILL.md 关联技法完整调取 / 0.0.1 and router 0.7 state that the
full result stays in receipts, the 本轮技法 panel and reports while web chat
answers from the card; computing the full spectrum is unchanged. Skill and
package version 6.9.16 -> 6.9.17 (tests/run_all.py 三栏).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_017eEAG8HD3mm8gsKXgk8uU8
This commit is contained in:
Jesse_Chen
2026-09-27 02:54:11 +08:00
co-authored by Claude Opus 5.5
parent 850f18605b
commit 147ebc1789
9 changed files with 42 additions and 23 deletions
+2 -2
View File
@@ -62,12 +62,12 @@ Every Level 2+ chart reading must include `MEVG / Global Web Evidence` and
For career, wealth, annual forecast, event timing, rectification, or a comprehensive reading, the system must **actually execute every available, question-relevant Indian and Western technique** before interpretation. A module merely existing in the repository, registry, or UI does not count as used.
1. Preserve every produced raw payload, varga, dasha, strength table, transit window, and cross-system difference. Information density must never be reduced because an oracle, formula variant, or holdout is incomplete.
1. Preserve every produced raw payload, varga, dasha, strength table, transit window, and cross-system difference. Information density must never be reduced because an oracle, formula variant, or holdout is incomplete. In web chat the full result is preserved in the server receipt, the folded 「本轮技法」 panel and the personal report; the answer-writing model reads the server's question-scoped evidence card (base chart plus the asked domain's section) and may look up one further section of the same calculation, read-only.
2. Invoke the Indian core relevant to the route: D1 plus all relevant vargas; Vimshottari, Narayana, Chara Dasha, Yogini, Ashtottari, Kalachakra and Bhrigu Pada where implemented; Jaimini/Arudha; KP; Double Transit PAC; transit LL/7L; Bhava Chalit; Sudarshana; Tajika/Varshaphala/Muntha; Nakshatra overlay; strengths, AV, yoga, argala, and relevant sahams.
3. Invoke the Western core for the same birth data: tropical natal chart, transit-to-natal duration scan, Solar Return, secondary progressions, solar arc, converse progressions/arcs, midpoints, Lunar Return, and parans where implemented.
4. A technique requiring unavailable inputs (for example partner data, Prashna question time, or an electional target) must still appear in the Technique Audit Table as `blocked` with its required input. It must not be silently skipped.
5. Keep raw information and conditional interpretations visible. The only restricted content is an unsupported certainty claim: no inevitable event, no unsupported exact day, and no claim of cross-system verification when only one system ran.
6. The report must state per technique: `executed`, `blocked`, or `not_applicable`, its settings, raw-output path, and its effect on assertion strength. `not_executed` is a workflow failure for a relevant technique.
6. The report must state per technique: `executed`, `blocked`, or `not_applicable`, its settings, raw-output path, and its effect on assertion strength. `not_executed` is a workflow failure for a relevant technique. In web chat that per-technique record is the receipt's folded 「本轮技法」 panel, not the answer body.
---
## 0.8 User-Led Reading Calibration Gate