Debrief: Turn Every Interview Round into a Revision Script

Where this fits

You just finished a technical interview. The call ended, the recruiter says "we'll be in touch," and you close the tab. But within 24 hours, the details start to blur — the exact question that tripped you, the interviewer's reaction to your system design answer, the hint they dropped about the next round. Instead of letting that fade, you want an agent that reconstructs the conversation, extracts what the interviewer cared about, and turns it into a concrete revision list you can use before the next call. That's the gap debrief fills.

Why agents benefit

• Reconstructs from messy input: parses a raw transcript, a freeform recollection, or a mix of both, marking uncertain items rather than fabricating answers.
• Guides memory recall: asks 2–3 questions at a time in chronological order, prompting the user to recall reaction details (follow-ups, note-taking, topic switches) without turning it into an interrogation.
• Produces a structured artifact: writes a rN-debrief.md with four fixed sections—Q&A reconstruction, interviewer profile updates, next-round forecasts, and a script revision checklist—ready for the agent to use in later rounds.
• Updates persistent intel: extracts interviewer style and disclosed team/project clues, feeding them back into an intel.md file that compounds across multiple rounds.
• Respects timing and scope: explicitly avoids mock-interview or pre-interview research use cases, keeping the skill focused on real post-interview reflection.

Practical scenarios

You paste a transcript from the call recording

Your agent reads the transcript, identifies each Q&A pair, notes interviewer reactions in parentheses, and flags where you were cut off or got stuck — then writes the debrief file.

You just say "刚面完" with no other context

The agent prompts you through the memory-recall sequence: first question, your answer, how the interviewer reacted, which ones you flubbed. It compiles your answers into the reconstruction, marking weak spots and suggesting specific script patches.

The interviewer hinted at the next round

The agent picks up on that hint (e.g., "next time we'll dive into distributed systems") and places it in the "下轮预测与建议" section, along with a forecast of likely topics based on the interviewer's style and your performance.

Add it to your agent workflow

Install the skill via the CLI:

npx skills add YunyueLi/greenroom --skill debrief

Then, in a conversation, say: "复盘一下今天的面经" or "转录贴在这里" — the agent will run the debrief routine and produce jobs/<slug>/rounds/r1-debrief.md. Use that file to generate a focused revision list for the next round.

Compare before adopting

Evaluate against three yardsticks:

• Quality signals: check the README for whether it handles uncertain memory explicitly (it does) and whether the output structure is stable enough for automation (it is, with fixed sections and frontmatter).
• Maintenance freshness: the project is early (v0.1.0), so watch for updates on the skill page and commit activity in the repo.
• Alternatives and workflow fit: if you already have a generic note-taking skill, compare how well it captures interview-specific intel versus this purpose-built one. This skill also pairs with mock-interview skills for pre- and post-practice loops.

Why it is worth tracking

This skill addresses a real pain point—post-interview knowledge decay—with a thoughtful, opinionated output format. At 11 stars, it's not mainstream, but the author's attention to recall psychology and the explicit anti-fabrication rule suggest deliberate design. Track it if you conduct multiple interview rounds and want your agent to compound learning round over round; re-evaluate when it gains community contributions or broader usage. The 24-hour deadline it imposes is a useful prod for your own agent to act fast.