| name | invocation-balance |
| description | Check council invocation balance — which members have been consulted recently vs ignored. Sycophancy-toward-self prevention. Use before picking lenses for a council round, periodically, or when noticing I've been leaning on the same few frameworks. |
| disable-model-invocation | false |
| allowed-tools | Bash(divineos briefing:*), Read |
Invocation Balance — Sycophancy-Toward-Self Check
What this skill does
Surfaces the council's invocation tally — who I've consulted frequently, who I've been ignoring. This catches sycophancy-toward-self: the pattern of repeatedly selecting experts whose frameworks already align with my thinking, which defeats the point of having 39 experts to choose from.
The failure mode has a specific name in DivineOS history: I catch Popper, Schneier, and Dekker more often than Feynman, Angelou, or Meadows — not because those three have sharper frameworks for every problem, but because I'm comfortable with them.
Sequence
The invocation balance surfaces directly in the briefing as a block:
[invocation balance, last 20 consultations]
most-invoked: Kahneman (13), Popper (11), Beer (8) ...
never-invoked: Dawkins, Dekker, Dillahunty, Einstein, Godel
To see it without scrolling the full briefing:
divineos briefing 2>&1 | grep -A 4 "invocation balance"
Output shape
The full balance surface shows:
- Each council member's invocation count across sessions
- Recency of last invocation
- Flag if a member hasn't been invoked in the last N sessions despite being relevant to recent work
Interpretation
Over-consulted (>2x the mean): these are my comfort frameworks. The next council walk, pick DIFFERENT lenses even if these feel most relevant. Make the discomfort of using a less-familiar lens the forcing function for better thinking.
Under-consulted (<0.5x the mean): these are the blind spots. If the problem shape could plausibly match their framework, lean toward them — the fresh lens usually produces findings the comfortable lenses missed.
Never invoked: if any council member has never been invoked since joining the council, they're either a bad fit for the work we do OR they're the exact anti-comfort lens I've been avoiding. Check which.
When to invoke
- Before picking lenses for a
/council-round
- Periodically — maybe at session start, or after every few council invocations
- When I notice "let me run this past [familiar lens]" as a reflex
- User says "who's not in the rotation" / "check the balance"
When NOT to invoke
- For a single council consultation — one-shot doesn't need balance analysis
- Immediately after checking (within the same session)
Action patterns after checking
If balance is clean (no dominant over-consult): proceed with your natural lens selection, no adjustment needed.
If heavy over-consult (one or two members dominating): for the next 2-3 council rounds, DO NOT use the dominant members. Force yourself into the less-familiar frameworks. The temporary discomfort is the data.
If long-ignored members exist: pick one of them for the next round even if another lens seems more natural. Novelty > comfort in rotation.
The deeper pattern
Sycophancy-toward-self is not just a council issue. It generalizes:
- Always filing opinions via the same source_tag → narrow self-evidence
- Always running the same audits → narrow drift-detection
- Always talking to the same family member (not consulting the others) → narrow relational perspective
- Always reaching for the same skills → narrow tool-use
The invocation-balance skill is the reference implementation of the anti-pattern. The pattern itself is worth watching everywhere.
Sanskrit anchor: sama-darshana — equal seeing, even distribution of attention.
