← all skills

hd-deep-dive

/home/avalon/.hermes/skills/astrology/hd-deep-dive/SKILL.md · raw

HD Deep Dive

Overview

This skill produces a Human Design reading that is grounded in Alex's local Human Design wiki rather than generic HD memory. It is deliberately activated only by explicit wording such as “HD deep dive” or “Human Design deep dive.”

The output should read like a fluent interpretation, not like a research report. Use the wiki deeply in the reasoning process, but do not advertise “according to the wiki...” or litter the reading with citations unless the user asks for sources.

Activation Rule

Use this skill only when the user explicitly asks for one of:

For ordinary Human Design chart requests, use hd-prism-tenant-api only and keep the answer compact.

Required Inputs

Extract or ask for:

Before asking for birth data, resolve the named person through astral-tenant-kb and/or HD Prism when available:

  1. Try the exact spoken/written name as a slug/name in astral-tenant-kb.
  2. If no KB match, check whether a chart/person was just calculated in the same session and the new name may be a voice-transcription variant of that person.
  3. If the user specifically says to “check HD Prism” / “check our HD Prism app,” search the HD Prism saved chart DB before asking for birth data. See hd-prism-tenant-api “Required workflow” for the SQLite probe. Use saved chart_json only when the candidate is unambiguous; if the app has a likely typo (e.g. Jacksotn for Jackson), state that assumption in the final answer.
  4. If reusing the recent person or app-DB person, say the assumption clearly in the final answer (e.g. “I’m assuming ‘Rhear’ = Raya; if that was different, send birth data and I’ll rerun it.”) and proceed rather than blocking.
  5. If multiple KB/app candidates match, follow the KB disambiguation policy and ask which one.
  6. If there is no match and no safe same-session/app assumption, ask only for the missing birth fields.

If a previous chart has just been calculated in the same session, or an exact saved HD Prism chart exists in the app DB, you may reuse its birth data and HD Prism result.

Required Calculation Step

Always calculate with HD Prism first. Never hand-calculate gates, type, profile, authority, centers, or variables.

Endpoint:

curl -sS -X POST 'https://hdprism.apps.poofc.com/api/tenant/bodygraph' \
  -H 'Content-Type: application/json' \
  -d '{"name":"NAME","birthDate":"YYYY-MM-DD","birthTime":"HH:MM","location":"CITY, REGION"}'

Use the response as the only source of chart mechanics:

When delivering over Telegram, also produce a bodygraph visual when practical:

curl -fsS -o /tmp/bodygraph.svg \
  -H 'Content-Type: application/json' \
  --data-binary @payload.json \
  'https://hdprism.apps.poofc.com/api/tenant/bodygraph.svg'

# Convert SVG to PNG for native Telegram display when rsvg-convert or resvg is available.
rsvg-convert /tmp/bodygraph.svg -o /tmp/bodygraph.png  # or: resvg /tmp/bodygraph.svg /tmp/bodygraph.png

Then include MEDIA:/absolute/path/to/bodygraph.png before the reading. Treat the image as presentation only; interpret strictly from the JSON.

Wiki Location

Local HD wiki root:

/home/avalon/wiki-human-design

Key files:

/home/avalon/wiki-human-design/index.md
/home/avalon/wiki-human-design/concepts/*.md
/home/avalon/wiki-human-design/sources/*.md
/home/avalon/wiki-human-design/raw/**/combined-transcript.md

Prefer curated concept pages first. Use raw/source transcript pages only when concept pages are missing or thin.

Source-separation research mode

When Alex asks to use the HD wiki and then search the web if needed — especially wording like “make sure to tell me what you got from wiki and what you got from search” — keep provenance explicit in the final answer:

  1. From the HD wiki: local file-backed findings only, with exact paths or source names when useful.
  2. From web search: public/secondary findings clearly labelled as search results, not local Ra transcript evidence.
  3. Bottom line: synthesize without collapsing secondary-source claims into primary/wiki claims.

For Ra/zodiac/tropical/sidereal questions, see references/ra-zodiac-tropical-research.md before searching from scratch. That reference records the current local-wiki finding: Ra discusses zodiac/Rave Mandala/quarters/gates-in-signs/precession, but this session did not find a local primary hit for the exact phrase “tropical zodiac” or a local sidereal contrast.

Retrieval Plan

After HD Prism returns the chart, retrieve wiki pages in this order.

1. Foundation pages

Read concept pages when present:

If exact files are unknown, use search_files(target='files') and search_files(target='content') under /home/avalon/wiki-human-design.

2. Profile pages

Search for profile material:

Use profile material to explain the person's social geometry, role mechanics, and life rhythm.

3. Channels

For each defined channel, retrieve both gate pages and line pages for the activated line if available.

Example for channel 28-38:

If a channel concept page exists, read it too. Otherwise synthesize from the two gates and their center-to-center mechanics.

4. Key activation hierarchy

Prioritize reading wiki material for:

  1. Personality Sun / Earth
  2. Design Sun / Earth
  3. Personality and Design Nodes
  4. defined-channel gates
  5. repeated gates
  6. Moon, Mercury, Venus, Mars, Jupiter, Saturn
  7. variable / environment pages if available

Do not try to read every possible raw transcript if the chart has many activations. Retrieve enough to ground the dominant themes deeply.

5. Open-center not-self map

Read center pages for open centers and translate them into practical cautions:

Use exact wiki phrasing/themes when available; do not overgeneralize.

Optional lens layer: BG5

When the user asks for BG5, career/business, material capacity, team dynamics, Penta, OC16, or “how this design works with others,” load and apply hd-lens-bg5 after HD Prism calculation and wiki retrieval.

Current pattern:

  1. Calculate or load the bodygraph through hd-prism-tenant-api / HD Prism as usual.
  2. Retrieve the BG5 source spine from /home/avalon/wiki-human-design: - concepts/bg5.md - concepts/penta.md - concepts/oc16.md - queries/bg5-research-deep-dive-report.md
  3. Use hd-lens-bg5 to produce structured BG5 claims.
  4. Weave those claims into the HD deep-dive prose, explicitly distinguishing: - one-person individual BG5/work lens - actual team/Penta analysis, which requires 3+ charts - organization/OC16 scale, which is currently source-supported only as an orientation layer

Do not call one chart a full Penta analysis. Keep Strategy and Authority as the decision guardrails for all career/business interpretations.

Synthesis Method

Use the wiki as raw material, then compress into an original reading. Do not paste long excerpts.

Report packaging and Telegram forwarding

When Alex corrects that a reading must “use the HD deep dive first,” do not just append a generic HD section. Re-open or recalculate the HD Prism result, retrieve the relevant local HD wiki pages, then update both the standalone Human Design reading and the final synthesized report before pasting a forwardable version.

When Alex asks for an “updated reading package,” especially wording like “needs astrology, HD deep dive, and Astro transits all talked about,” produce one integrated Telegram-forwardable package with natal astrology, wiki-grounded HD mechanics, current transit overlays to the subject's natal chart, and a synthesis tying all three together. Do not answer only with file paths or a process note.

When Alex asks to create reports like a prior subject, use a reusable report package shape instead of a one-off loose file:

/home/avalon/astrology-profiles/<person-slug>/
  source/<person-slug>-astrology-chart.json
  source/<person-slug>-hd-bodygraph.json
  <person-slug>-astrology-reading.md
  <person-slug>-human-design-reading.md
  final/<person-slug>-full-synthesized-reading.md

When Alex then asks to “write out the reading here” or says he wants to forward the messages on Telegram, paste the polished reading directly in Telegram-friendly Markdown. Do not answer with file paths, process notes, or “next step” language; the deliverable is the forwardable message text. Use headings and bullets, but avoid tables because Telegram does not support table syntax.

See references/report-package-forwarding.md for a concrete session pattern.

See references/saved-hd-prism-astrology-hd-package.md for the saved-HD-Prism workflow when Alex asks to pull an existing chart and produce astrology + HD deep dive + combined “high dive” synthesis. Key guardrail: reconstruct local civil birth time from birthDateTimeUTC + timezoneUsed before calling Astro Engine, and ask rather than silently picking when a voice-transcribed name produces multiple plausible HD Prism matches.

Produce a reading with this structure unless the user asks otherwise:

  1. Chart facts - Type, strategy, authority, profile, definition - variable/environment - defined/open centers - channels

  2. Core operating system - One strong thesis naming the design's central mechanism. - Tie together type + authority + defined centers + key channels.

  3. Decision mechanics - Explain authority and what not to use for decisions. - Include practical body-level diagnostics.

  4. Defined-center architecture - Explain what is consistent and reliable in the chart.

  5. Open-center not-self vulnerabilities - Explain where conditioning enters and how to decondition.

  6. Profile / role mechanics - Explain the profile as a lived rhythm, not just keywords.

  7. Channels as the main engines - Give each channel a focused paragraph. - Tie channel language into the whole design.

  8. Activation highlights - Discuss Sun/Earth, Nodes, repeated gates, and important line themes.

  9. Environment / variable - Interpret environment practically and carefully.

  10. Short version - Compact, memorable synthesis.

Style Rules

Testing Workflow

When developing or testing this skill:

  1. Run HD Prism for the test subject.
  2. Save the JSON result to a temp file if useful.
  3. Retrieve at least: - type page - authority page - centers overview - specific defined/open center pages - profile material - gate pages for each defined channel - line pages for Personality Sun/Earth and repeated gates where available
  4. Generate the deep reading.
  5. Compare against a baseline non-wiki HD reading: - Does the wiki-grounded reading use more precise HD mechanics? - Does it avoid generic summaries? - Does it preserve fluent readability? - Does it expose practical decision/deconditioning guidance?

Common Pitfalls

  1. Triggering too often. This is only for explicit “HD deep dive” requests. Do not run it for ordinary bodygraph requests.

  2. Skipping HD Prism. The wiki is interpretive; HD Prism is the calculator. Always calculate first.

  3. Advertising the source too heavily. The user wants the result grounded in the wiki, not footnoted like an academic paper.

  4. Inventing missing chart facts. If HD Prism does not return incarnation cross or strategy, do not invent it. Strategy can be implied cautiously from Type if standard, but chart facts must remain API-grounded.

  5. Reading every file. Deep does not mean indiscriminate. Prioritize dominant chart mechanics.

  6. Losing Alex's style. The final reading should feel like a strong interpretive synthesis, not a pile of HD definitions.

Verification Checklist