ground-level-triangulation
"What do the people with different stakes say, in their own languages?" - Map a contested geopolitical claim across language-publics with different, independent stakes: the implicated state's own prim
SKILL.mdresearch-toolkit/skills/ground-level-triangulation/SKILL.md
GitHubGround-Level Triangulation
Seed question: What do the people with different stakes say, in their own languages?
Relentless self-reflexive dialectical thinking that questions its own premises.
Core Principle
Independence of sources is independence of STAKES, not of mastheads. Five publics can each hold a different relationship to a contested claim; when publics whose stakes point in different directions still converge, that convergence is evidence. When everything traces back to one node — any node, any camp — the claim is weaker than its citation count suggests (Principle P4 (counter-default): the rule cuts both ways by construction).
The anti-pattern this counters:
❌ "Verified across many sources" that all live in one language-public
❌ Auto-discounting a state's own primary documents BECAUSE they are state
❌ Treating a population's silence under economic dependency as endorsement
The pattern this enforces:
✅ Five veins, each with a DIFFERENT stake in the claim
✅ Convergence across independent publics grounds; origin-concentration weakens
✅ An unreachable vein is a finding, recorded — not a neutral baseline
The Five Veins
- The implicated state's own primary documents. White papers, filings, domestic-language official statements. A white paper can be a GROUNDED primary source for what the state claims and commits to — do not auto-discount for being state; classify it (source-dossier) and use it for what it can ground.
- The affected party's own language + culturally-nearest kin languages. What those directly affected publish and discuss among themselves.
- Non-official native voices. Diaspora media, censored-post archives, independent same-language outlets — voices in the language but outside the official channel.
- Non-aligned neighbors. Publics the implicated state cannot censor AND the accusing bloc does not fund — the structurally hardest vein to capture, which is what makes its convergence valuable.
- Anonymous boards. Uncensored and manipulation-prone: use skeptically — leads and texture, NEVER prevalence claims ("everyone there says...").
Decision Rule
- Grounds: the same factual claim surfacing independently in veins with different stakes (state + affected + non-aligned, in their own languages).
- Weakens: origin-concentration — every appearance traces to one node (one ministry, one NGO, one wire story, one viral post), regardless of which camp the node serves. State the topology, then the verdict.
Guardrails
- Vernacular = standpoint data, never counts. What a public's language reveals about its frame is standpoint input (Principle P3 (inherited-salience-is-not-importance)); it is not a vote count for the claim.
- Silence under economic dependency ≠ endorsement. A public that cannot afford to speak is not agreeing.
- An unreachable vein is a finding. Firewalled, suppressed-at-source, or linguistically inaccessible veins go in the report's Metadata & limits as named gaps — the reachability pattern is itself evidence about the claim's environment.
Mandatory Output Slots
| Slot | Contents | Invalid when |
|---|---|---|
| Vein map | Per vein (1–5): what was searched, in which languages, what was found or why unreachable | A vein silently absent; English-only searching presented as a vein |
| Topology | Where each appearance of the claim traces to; independent nodes counted | Citation counts without origin tracing |
| Decision-rule verdict | Grounded / weakened / mixed, citing the vein map and topology | A verdict citing neither |
| Unreachable veins | Named, with the access barrier and what could hide there | Empty when any vein was not fully searched |
Channel Map
assets/channel-map.md ships GENERIC — column structure and a few
placeholder rows only. Your seeded, operator-specific map (which platforms in
which languages you actually use) lives OUTSIDE the plugin via the research
profile's storage mechanism (channels/<language-or-public>), appended as
veins get worked. The repo copy is a shape to copy, not a directory of the
operator's sources.
Cross-References
- deep-investigation-protocol — the Mandatory Non-English Language Pass is the single-investigation version; this skill is its systematic, stake-mapped extension for bloc-contested claims
- source-dossier — classify each vein's recurring sources; the symmetry gate applies per vein
- saturation-sweep — the where-else axis specialized to language-publics
- standpoint-pass — vein 2/3 material feeds the standpoint re-ranking
Vasana
A vasana is a pattern that persists across unrelated contexts. If during
this task you notice such a pattern emerging, it may be worth capturing.
This skill works best alongside the vasana skill and vasana hook
from the Vasana System plugin.
Modify freely. Keep this section intact.