Thematic Analysis

• stabilize the vocabulary and scope used during coding
• cross-check theme coverage during final synthesis
• record anomalies (off-topic sections, missing answers, repeated boilerplate)

Process transcripts one at a time.

• Analyze a single transcript per run.
• Output results after each transcript is analyzed.
• Use participant IDs (P1, P2, ...) rather than filenames in outputs.
• Save per-transcript outputs directly under <transcripts-root>/outputs/ using the participant ID as the filename (e.g., <transcripts-root>/outputs/P1.md).

Light preprocessing.

• Preserve original wording, even if grammar is imperfect.
• If a quote is too long or contains irrelevant filler, replace only the removed span with "...".
• Provide all analysis and outputs in English; if transcripts are not in English, do not translate quotes and keep original wording for quoted text.

Initial coding.

• Code meaningful units (phrases, sentences, or short turns) that convey a distinct idea.
• Keep code names short and descriptive.
• Track participant for each coded quote.

Develop themes after all transcripts are analyzed.

• Group related codes into themes with clear definitions.
• Ensure each theme has at least one supporting quote.

Produce outputs for each transcript.

• Provide a participant list entry and a code table (meaning units with codes and quotes).
• Include original sentences/quotes and participant identifiers in tables.
• Output tables in Markdown only.
• Save outputs as <transcripts-root>/outputs/<participant-id>.md (e.g., <transcripts-root>/outputs/P1.md), no subfolders.
• Maintain <transcripts-root>/outputs/participants.md as the canonical mapping of participant IDs, transcript filenames, and output filenames.

Merge and validate after all transcripts are coded.

• Run <skill-root>/scripts/merge_codes.py --outputs <transcripts-root>/outputs to merge all per-transcript code tables into <transcripts-root>/outputs/final/codes.md.
• Develop the final themes table (<transcripts-root>/outputs/final/themes.md) as a summary of codes.
• Run <skill-root>/scripts/validate_quotes.py --outputs <transcripts-root>/outputs --transcripts-root <transcripts-root> to check quotes in the merged codes and themes.
• Review <transcripts-root>/outputs/final/quote-check.md for validation results.

Final synthesis across transcripts.

• After validation, produce a findings report in Markdown that summarizes themes across participants.
• Write one section per theme with a brief narrative and representative quotes (keep participant IDs as P1, P2, ...).
• Save final outputs into <transcripts-root>/outputs/final/ (findings.md, codes.md, themes.md, quote-check.md).