skill: persona-based research framing (Islamic scholar / historian / media studies), informed by local scholars & imams

[fmadore]

Summary

Today the iwac-mcp skill branches research on a single axis: depth — the Brief vs. Extended choice (a how much decision about effort and tokens). It does not branch on framing — which questions matter, which vocabulary to search, what counts as a relevant finding, and how to weight the evidence. Those are exactly the things that differ between disciplines and, more importantly, between an outside academic and someone inside the Muslim community.

This issue proposes adding a persona layer: the user (or Claude) picks a research lens that reshapes the five-phase workflow, instead of — or alongside — the Brief/Extended depth choice.

Motivation

Two researchers can ask "what does IWAC say about X?" and want genuinely different answers:

• A historian wants chronology, change over time, the 1990–91 press-system break, periodisation, named actors and institutions traced across decades, source criticism.
• An Islamic studies scholar wants doctrine and jurisprudence, Sufi orders / tariqa, reformist vs. Sufi vs. Salafi/Izala currents, religious authority (imams, ulama, arabisants), religious education, intra-Muslim debate.
• A media / communication studies scholar wants framing and representation, editorial lines, who speaks vs. who is spoken about, agenda-setting, sentiment, the francophone press's gaze on Muslims.

The current skill bakes in one implicit lens (close to the historian's, since that's the project's default voice). Depth tuning doesn't change what the research treats as important; only a framing layer does.

Crucially, this connects to the skill's existing bias ethos. The skill already discloses that the corpus over-represents Western-educated, French-speaking Muslims and under-represents arabisants (SKILL.md, "Key Constraints" §2). The persona idea extends that honesty from the data to the research framing: the questions we ask are themselves a perspective, and we should be able to choose — and name — which one we're using.

Proposal: research personas as a framing axis

A persona is a preset that tunes each phase of the existing workflow:

Phase	What the persona changes
1 Scoping	which subsets to prioritise (e.g. references + index for the scholar; articles + sentiment for media studies; everything chronologically for the historian)
2 Search	which vocabulary to lead with (doctrinal terms vs. event/actor names vs. framing/representation terms) and which transliteration variants matter
3 Deep reading	what to read closely (a fatwa or doctrinal passage vs. a datable event vs. an editorial framing)
4 Triangulation	what to cross-check against what
5 Synthesis	how findings are organised and what confidence means in that discipline

Candidate personas

• Historian — change over time, periodisation, actors/institutions, the press-system break.
• Islamic studies / religious studies scholar — doctrine, jurisprudence, currents, religious authority, education.
• Media / communication studies scholar — framing, representation, discourse, sentiment, editorial lines.
• (later) Social scientist / anthropologist — practices, community life, networks.

The important part: community-informed relevance

The personas above are all outsider academic lenses. The original idea goes further: let local scholars or imams help define what is important about the Muslim community in the first place — so the skill doesn't only encode what external researchers think matters.

Concretely this could be:

• a "community insider / local religious authority" persona whose priorities (the religious calendar and its observance, da'wa, mosque and association life, Quranic schooling, zakat/charity, pilgrimage, local religious authority and succession, intra-community concord) are sourced from consultation with local scholars and imams rather than from disciplinary categories; and/or
• a consultation step in building the skill: interview local scholars/imams about what they would want surfaced from a corpus like this, and fold that into research-domains.md and the persona definitions.

This is the most novel and most valuable piece. It is a participatory / partly decolonial move: it treats what counts as important about the Muslim community as something to be defined with the community, not only for it. It should be flagged as needing real input (not something Claude can fully invent), and it pairs naturally with the project's existing under-representation disclosures.

Open design questions

1. Replace or augment? The phrasing was "instead of having two research processes [Brief/Extended]." Should persona replace depth, or be an orthogonal axis (persona = what/how, depth = how much)? My lean: orthogonal — keep Brief/Extended for budget, add persona for framing — but this is the central decision to settle first.
2. Who picks, and when? Always prompt for persona? Infer it from the question and confirm? Default to a "generalist" persona that behaves like today's skill?
3. One persona or a blend? Real questions are often cross-disciplinary; should personas be composable (e.g. "historian + media studies")?
4. How to genuinely source the community persona without it becoming Claude role-playing an imam — i.e. how to keep it grounded in real consultation and clearly labelled as a framing aid, not an authoritative religious voice.
5. Avoiding caricature. Personas must not flatten disciplines into stereotypes; each needs a careful, honest definition.

Sketch of where it would land

• New references/personas.md defining each persona's phase-by-phase emphases (loaded on demand, like the other reference files).
• A short persona-selection section in SKILL.md, parallel to (or merged with) the current "Research Depth" section.
• Persona-specific search-term emphases threaded into references/research-domains.md.
• Possibly a brief methods note in the synthesis output naming the lens used ("framed as a media-studies reading"), in the same spirit as the existing evidence-base ledger.

Related

• skill: make .agents/skills/iwac-mcp model-agnostic (Qwen / local LLM support) #4 (skill: make iwac-mcp model-agnostic) — touches the same SKILL.md; worth keeping the persona presets simple enough to survive on smaller/local models too.