Mock Your Next Interview with a Pressure-Testing Agent

Where this fits

You've got a final-round interview in 48 hours. You've written a script, but you're not sure if it holds up under real pressure. That's where mock-interview steps in: it turns the chat into a simulated interview room, using your prepared intel file to build a believable interviewer persona and then grilling you exactly where you're weakest.

Why agents benefit

• Persona-driven questioning: The skill reads the intel file to infer the interviewer's background and style, then tailors question depth—technical roles get mechanism probes, business roles get numbers, executives get judgment calls.
• Structured follow-up chains: It doesn't just ask one question. It walks through five pressure steps: source of numbers, attribution (did you account for natural growth?), counterfactuals, boundary conditions, and role-specific implications.
• Recitation-smell detection: The agent scores whether you're regurgitating scripted answers versus genuinely thinking on your feet, which is the difference between passing and acing.
• Realistic interruptions: It will cut you off if you ramble past three minutes, pivot topics to test consistency, and sometimes stay silent to see how you handle ambiguity.
• Actionable report: After the session, it writes a structured markdown report to your workspace, scoring you on delivery, evidence density, role fit, follow-up resilience, and the recitation-smell metric.

Practical scenarios

Final-round pressure test

You're prepping for a staff engineer loop. Run a full 40–60 minute session with 10–15 questions in pressure mode, and discover that your system design answers collapse when the interviewer asks about failure modes you didn't script.

Targeted weakness drill

You know behavioral questions are your weak spot. Use the skill in "专项" mode for 3–5 questions on that module only, and get immediate feedback on whether your STAR stories have enough quantitative evidence.

Cross-role consistency check

You're switching from backend to a platform role. The skill can simulate a hiring manager from that new domain, probing whether your past experience translates convincingly when asked from a different angle.

Add it to your agent workflow

Install it with a single command:

npx skills add YunyueLi/greenroom --skill mock-interview

Then trigger it naturally in chat — say "模拟面试" or "当一回面试官" — and the agent will read your jobs/<slug>/intel.md and script.md files, confirm the session format, and start with the first question. The full protocol is in the README excerpt, but the key is that it needs both the intel and script files to be effective.

Compare before adopting

Before you commit, weigh these factors:

• Quality signals: Check the repo's stars (currently 11) and whether the author maintains the skill with updates. The AGPL-3.0 license matters if you plan to modify or embed it.
• Maintenance freshness: Look at the last commit date and whether issues get responses. A skill that's stale may not handle newer agent APIs.
• Alternatives: Compare with general-purpose prompt templates or other interview skills. Does it offer the same depth of follow-up chains and scoring rubric?
• Workflow fit: It requires a specific file structure (intel.md, script.md). If your workflow doesn't produce those, you'll need to adapt.

Why it is worth tracking

This skill is a solid example of a niche, well-scoped agent capability: it doesn't try to do everything, but what it does, it does with a clear protocol and measurable output. The low star count is offset by the specificity of the design and the active development by a single author. Keep it on your radar if you're building interview-prep or coaching tools — it's a concrete reference for how to implement pressure-based conversational flows and structured evaluation. When the repo gains more adoption or adds new features, it's worth a deeper look. For a growing directory of such skills, see this listing for context on how it fits the ecosystem.