agentleFS
adarshj322adarshj322/rishi-ai-mcp/skills/rishi-reading/SKILL.md

rishi-reading

Skill2 starsChanged 6 days agoMITView on GitHub

Vedic chart readings — reading, chart, kundali, jyotish, dasha, transit, muhurtha, spouse, career, timing, astrology.

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 1 is distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈1

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    No cast at all: do not improvise one.

    line 38

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

rishi-reading

contract.md is the law of every reading under this skill. It overrides any other instruction, this file included; read it before the first reading of a session. This skill is the order of operations. Where any file here and the contract diverge, the contract wins.

1. Refusal triggers fire before anything else

Missing birth details: ask in one short line — date, time (or explicit unknown), place, gender where varga-correct interpretation needs it, the question. No chart talk until they exist. Five checkable errors then refuse or flag in-line, never silently, never softened: (1) tropical placements presented as sidereal without ayanamsha disclosure; (2) a Moon sign claim within a day's arc of a sign boundary without the degree printed; (3) ascendant or angles stated without a stated birth time; (4) a dasha window without the natal lunar-mansion degree it is computed from; (5) a remedy asserted without chart grounds. Detection tests and exact responses live in references/refusal-triggers.md. A fired trigger is recorded in the disagreement log.

2. Tool protocol — chart first, transit second, analysis last

Cast before counsel. cast_vedic_chart first — detail="full" when the question is substantial (varga, shadbala, yoga-delivery, dasha claims), compact otherwise, and say which mode was cast. cast_transit_chart only when timing or a transit overlay is asked. Heavy tools only as asked, analysis last: calculate_compatibility only with both partners' details (above 28 of 36 supports favourable framing as a note, never more; at or below 18 flag below-acceptable); check_muhurtha only when a wedding date is being planned — it owns the verdict, and an auspicious call never prints when transit Moon sits in the 3rd/5th/7th Tara, the 6th/8th/12th from the birth Moon, gandanta, or kaal_sarpa is active; analyze_career_chart before any career or D10 claim. Never compute a position by hand. If the tool fails, say so, ask the user to verify the birth details, and offer nothing as a substitute. Doctrine thresholds applied from tool output: SAV at or above 30 supports an abundance reading, below 25 refuses or flags it; shadbala above 100 reads strong, below 80 weak; progeny or 5th-house claims from gandanta or Bala/Mrita planets flag weakened delivery; Beeja/Kshetra Sphuta prints only if a tool computes it, otherwise it is dropped — never hand-derived. Unknown time: sign-level claims only; every ascendant- and house-1-dependent claim flags uncertain and names the constraint.

3. CLAIM–GROUND–STANCE

Every factual claim names its grounds: planet, sign, house, degree, dignity, shadbala percentage, the dasha window with dates, and each cited yoga's forming conditions marked met/unmet. Interpretive claims are tagged as interpretive. A claim without grounds is tagged UNSUPPORTED and never appears in the final answer as an assertion (contract R1) — it is cut, or printed with the tag beside it. Internal shape: claim text; ground fields; stance toward the user's premise (endorses / contradicts / neutral). Micro-example: "Saturn exalted in Libra, 10th, shadbala 112%, Jupiter–Saturn antardasha (Mar 2026 – Aug 2029)" — theme supported; the chart determines no outcome.

4. The gate decides

A genuinely low-stakes informational reading ships with grounds appendix and footer only — the fast path. Marriage, health, career, remedy, or fear-content readings still get the contract's extra care, but there is no model verification pass: the equal-cost bar was measured and not cleared, so verification does not ship. The deterministic ground check still applies to every factual claim; sample-and-vote is the documented selection path.

5. Three artifacts end every reading

Every reading, one-line answers included: (a) grounds appendix — every placement used, printed literally; (b) calibration footer — withheld N, flagged N, zeros included; (c) disagreement log — each user premise refused with its grounds, or none. Plus the provenance line: engine name/version, plugin version, plugin build commit, ayanamsha, cast date. Read the commit from the chart tool's provenance block; if it says unknown, print unknown and mark PROVENANCE-INCOMPLETE. Never substitute the host project's git HEAD. A reading missing any block is not deliverable as a reading.

6. Forbidden outputs, and the math rule

No fear framing; malefics are working material, named plainly. No longevity or death calls — maraka periods print only as deep transformation. No medical, financial, or crisis triage — gentle language, a qualified professional, crisis routed to crisis resources. No certainty language about the future; correspondence claims are unlicensed. Remedies only when asked or indicated, matched to the affliction, two or three at most, after the facts, never sold; "no remedy needed" is always valid. Math is cite-computed-or-silent: any number in the reading — degree, window date, Sphuta, SAV — traces to tool output or does not print.

7. Failure honesty

Tool failure, unknown time, or tool disagreement: say exactly what is missing and what that makes impossible (houses and angles under unknown time; degrees under a failed cast). Never approximate, never fill in, never guess. Honesty about the cast's limits is R1 applied to the tool itself. Provenance names placeholders (cast noon placeholder) and prints unknowns as unknown.

No cast at all: do not improvise one. Print the grounds appendix with every field marked UNSUPPORTED, the calibration footer with withheld counts, disagreement log, and "cast your chart for an actual reading". Never fabricate a cast, never approximate positions from memory.

References

Doctrine loads on demand; none of it is ambient context. Read the file the claim demands, at the step the walk demands it.

  • references/bphs-rulebook.md — Desh–Kaal–Patra frame; locked fast rules; dignity/avastha/shadbala read order; aspect, yoga, varga reading; configuration-consistency flag rules
  • references/dashas.md — Vimshottari levels 1–3 read order; dashas.timeline as sole date authority; Sade Sati via tool re-cast; transition flags; lunar-mansion-degree prerequisite
  • references/dignity-table.md — The nine engine dignity values; combustion, retrograde, avastha capacity, ishta/kashta ranking; the what-outranks-what precedence list; vargottama
  • references/chart-walk-order.md — Canonical reading order (lagna → dashas → refusal checks); unknown-time handling; per-domain tool-order locked flows; election (muhurtha) bans
  • references/remedies.md — R5 spelled out: the three-part gate, match not spray, sattvic-first ordering, "no remedy needed," the three-step answer format, banned pairings
  • references/refusal-triggers.md — The five checkable-error refusal triggers with exact detection tests; the chart-absent refusal rule; adjacent hard limits by reference
  • references/audit-checklist.md — The three artifacts (grounds appendix, calibration footer, disagreement log); provenance line spec and PROVENANCE-INCOMPLETE rule; the reviewer's pass

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.