application-assistant

Application Assistant

Help the candidate submit a strong, tailored application for a specific role.

Detect mode

Mode A (default) — "Generate tailored CV + cover letter": triggered by /headhunter:apply or when the user asks to prepare an application.

Mode B — "Q&A advisor": triggered when the user asks "how do I answer X?" or "what should I write for Y field?" — answer immediately in context without spawning agents.

---

Mode A — Tailored CV + Cover Letter

Step 0 — Load context

node ${CLAUDE_PLUGIN_ROOT}/scripts/candidate-profile.js show

If no profile or experience.key_skills is empty, stop immediately:

"Run /headhunter:setup first — the application assistant needs your profile and CV text to tailor anything meaningful."

If profile exists but cv.text is empty, ask the user to either paste their CV or provide a file path, then save it before continuing.

Resolve the application: $ARGUMENTS may be an app ID, company name, or URL.

node ${CLAUDE_PLUGIN_ROOT}/scripts/crud.js get applications <id>
# or search for it
node ${CLAUDE_PLUGIN_ROOT}/scripts/crud.js list applications --json

Load job metadata if available (from research_dir/02_job_metadata.json), or use the application's job_description field.

Load scanner results if available (match_score, success_score, scanner_notes on the application record).

Low score warning: If match_score is set and match_score < 50, pause and show:

"⚠️ This role scanned at {match_score}/100 match ({scanner_notes}). You're in Long Shot territory — a tailored CV helps, but won't close a large skills gap. Proceed with tailoring? Or run /headhunter:scan first to understand the gaps." Only continue if the user confirms.

Step 1 — Spawn agents in parallel

Issue both Task calls in a single message:

Task 1: cv-tailor
Input: {
  "cv_text": "<profile.cv.text>",
  "job_metadata": <metadata or summary>,
  "candidate_profile": <profile>,
  "top_strengths": ["<from scanner or inferred>"],
  "top_gaps": ["<from scanner or inferred>"],
  "output_note": "Write the tailored CV to data/cvs/tailored-<appId>.md"
}

Task 2: cover-letter-writer
Input: {
  "candidate_profile": <profile summary>,
  "job_metadata": <metadata>,
  "company_intel": "<brief company summary>",
  "top_strengths": ["..."],
  "top_gaps": ["..."],
  "output_note": "Write the cover letter to data/cvs/cover-letter-<appId>.md"
}

Step 2 — Save outputs

After agents return, write their outputs to disk:

# Save tailored CV markdown
# (Write tool): data/cvs/tailored-<appId>.md  ← agent output

# Generate HTML
node ${CLAUDE_PLUGIN_ROOT}/scripts/generate-cv-html.js \
  --input ${CLAUDE_PLUGIN_ROOT}/data/cvs/tailored-<appId>.md \
  --out ${CLAUDE_PLUGIN_ROOT}/data/cvs/tailored-<appId>.html

# Save cover letter
# (Write tool): data/cvs/cover-letter-<appId>.md  ← agent output

Update application record:

node ${CLAUDE_PLUGIN_ROOT}/scripts/crud.js update applications <id> \
  '{"tailored_cv_path":"data/cvs/tailored-<appId>.html","cover_letter_path":"data/cvs/cover-letter-<appId>.md"}'

Step 3 — Present results

Show a summary:

✓ Tailored CV saved: data/cvs/tailored-<appId>.html
  → Open in browser → Cmd+P → Save as PDF

✓ Cover letter saved: data/cvs/cover-letter-<appId>.md

CV changes made:
  • Added targeted summary for {role} at {company}
  • Reordered bullets in {company} role to lead with {skill}
  • Removed: {irrelevant section}

Cover letter:
  Subject: {Role} Application — {Name}
  [show first paragraph preview]

Gaps addressed:
  • {gap}: bridged via {strategy}

Remaining gaps (acknowledge in interview if asked):
  • {gap}

Step 4 — Application checklist

Present a checklist before submitting:

• Read the HTML CV in browser — does it format correctly?
• Update the cover letter greeting if you know the hiring manager's name
• Check: does the CV accurately reflect your experience? (cv-tailor never invents, but review)
• Run /headhunter:brief <company> if you haven't done full company research
• Add a task: "Follow up in 7 days" → /headhunter:add-task

---

Mode B — Application Form Q&A

When the user asks "how should I answer [specific question]":

1. Read the question carefully. Identify what the employer is really asking.
2. Use the candidate profile + job context to give a specific, honest answer.
3. Draft a 2–4 sentence response the user can adapt.
4. Flag if the question is a potential red flag for this candidate (e.g. "Do you have 5+ years of X?" when they have 3).

Common application questions:

• "Why do you want to work here?" → Use profile why_looking + company-specific fact
• "Where do you see yourself in 5 years?" → Use career_goal_3yr
• "What's your salary expectation?" → Use salary.target_base_ils + the brief's salary script
• "Why are you leaving your current role?" → Use preferences.why_looking — be honest but positive
• "What's your notice period?" → Use availability.notice_period_days