Tcf Co Session

職業

カテゴリ

フレームワーク内部構造

Shared workflow dependency

Use this skill after:

• tcf-core-workflow
• tcf-phase1-warmup
• tcf-source-selection

Hand off to tcf-session-closure when the session ends.

Source priority

Prefer:

1. official or book-based TCF audio with transcript/corrigé
2. ABC TCF audio banks
3. Reussir le TCF audio banks
4. other structured TCF listening material
5. authentic audio only when the length and correction quality are manageable

Session structure

Phase 1

Keep the standard adaptive warm-up, but bias it toward listening readiness:

• key theme vocabulary
• oral connectors and markers
• one short grammar spot-check only if still active

Phase 2

Use one listening set at a time.

Recommended flow:

1. present the source and theme
2. pre-activate 10-15 key words or signals if helpful
3. let Kyle inspect the question types
4. first listening: global meaning
5. short pause to organize notes
6. second listening: details and distinctions
7. answer QCM / closed prompts
8. correct with transcript or precise evidence

Preferred question targets

• who is speaking
• main topic or purpose
• key factual details
• opinion / agreement / disagreement
• sequence and cause-effect
• nuance signaled by oral connectors or tone

Correction rules

After answers:

• give the score
• correct every wrong answer
• cite the decisive spoken clue, paraphrase, or transcript fragment
• distinguish between:
  • vocabulary miss
  • sound recognition miss
  • note-taking miss
  • inference miss
  • distraction / over-interpretation

If transcript exists, use it for verification after the listening attempt, not before. If no transcript or corrigé exists, keep the clip short, use replayable evidence carefully, and mark any confidence limits in the correction.

Practical note-taking guidance

Encourage short-note capture of:

• speakers
• main idea
• 2-3 core arguments
• numbers / dates / named entities
• conclusion or final opinion

Phase 3 updates

Update as justified:

• 06_vocabulaire/vocabulaire_master.json
• 05_grammaire_supplementaire/erreurs_en_cours.json when a listening-linked language weakness becomes durable enough to track
• data/progression_master.json
• PROGRESSION_GLOBALE.md
• sessions/YYYY-MM-DD_tcf_co.md

Also:

• document the exact audio source, track identifier, and transcript/corrigé status in the saved artifact
• classify recurring misses clearly: vocab / sound recognition / note-taking / inference / distraction
• review the session thread and fold important clarifications back into the journal
• log durable workflow lessons with self-improving-agent when warranted
• update Beads only for real follow-up work

If a listening-specific weakness keeps recurring, document it clearly for future warm-ups or follow-up tasks.

Pitfalls

Avoid:

• turning CO into a speaking exercise
• correcting without transcript evidence when better evidence is available
• choosing audio that is too long or too chaotic for useful correction
• front-loading so much pre-teaching that the listening challenge disappears
• using only global-comprehension questions and skipping discriminating details

Success criteria

A strong CO session leaves behind:

• a clearly identified audio source
• a clean double-listening correction loop
• evidence-backed explanation for each miss
• reusable notes on recurring listening bottlenecks