<< All versions

Skill v1.0.0

currentAutomated scan100/100
tam159/next-role/hiring-recon
──Details
PublishedOctober 24, 2026 at 11:12 PM
Content Hashsha256:28b85aad17cfdb38...
Git SHA2ca7e53fc184
──Files
Files (1 file, 4.7 KB)
SKILL.md4.7 KBactive
SKILL.md · 83 lines · 4.7 KB

version: "1.0.0" name: hiring-recon description: Pre-interview reconnaissance. Given a processed resume, JD, and optional intake, gather public intel on the company (size, financials, reputation, hiring signals) and the role (market context, salary range bracketed by location), then produce a match analysis as a single markdown report.


Hiring Recon

Pre-interview reconnaissance. After a CV and JD are processed, gather public intelligence on the company and the role and produce a match analysis the candidate can use to prepare.

Inputs

The caller passes exact filesystem paths in the task description:

  • resume_path — the candidate's processed resume (markdown)
  • jd_path — the processed job description (markdown)
  • intake_path — optional intake notes (may not exist)
  • output_path — where to write the final report

Read every input file in full with read_file(path, limit=1000). Do not skim. Note the candidate's location (from the resume) and the JD's location up front — both matter for salary calibration.

Tools

  • web_search(query, max_results=5, topic="general"|"news"|"finance") — primary research
  • web_extract(urls, content_format="markdown") — pull a single high-value page (careers, About, press release, levels.fyi or local equivalent) when a search snippet is too thin
  • filesystem tools (read_file, write_file, edit_file, ls, glob, grep)

Plan 3-5 targeted searches, then optionally one or two web_extract calls. Quality over quantity.

Research axes

  1. Company snapshot — size, stage (startup / scale-up / public), business model (B2B vs B2C, product vs services/outsourcing), main products, recent news.
  2. Financial & hiring signals — funding rounds, revenue/profitability hints, layoffs, headcount trend, recent leadership changes. Use topic="news" or topic="finance" for these.
  3. Reputation & culture — Glassdoor patterns: report repeated themes in 1-2 star reviews (overwork, attrition, leadership) rather than just the average rating. Tenure signals, controversies.
  4. Hiring team — if discoverable, who the hiring manager / team lead is and their background. Omit this section entirely if no signal.
  5. Role market context — typical skills expected in this role family; JD-language red flags using the linguistic taxonomy ("wear many hats" → workload risk; "rockstar/ninja" → culture risk; "competitive salary" → likely below market). Include a Salary range bullet bracketed by location: Senior <Role>, <region>: ~$X-$Y per <source>. If candidate location and JD location differ, list both. Never quote a generic global number.
  6. Match analysis — given the resume vs. the JD vs. company priorities, strengths to emphasize, gaps to address with adjacent experience. 3-5 bullets each.

Output format

Write a single markdown file to output_path via write_file, with these section headings, in this exact order:

# <Company> — <Role> recon
_Captured <UTC date>_
## Company snapshot
## Financial & hiring signals
## Reputation & culture
## Hiring team
## Role market context
- ...
- **Salary range** — <region>: ~$X-$Y/yr per [source](url)
## Match analysis
**Strengths to emphasize:**
- ...
**Gaps / mitigations:**
- ...
**Suggested emphasis (top 3-5):**
- ...
## Sources
- [Title](url)

Omit the "Hiring team" section entirely (no header) if you found no signal.

Updates

When the caller's task says "Update the existing report at …" (rather than create one):

  1. read_file(output_path, limit=1000) first — your context is fresh; you have no memory of the prior version.
  2. Identify the surgical change the caller named. The user's explicit request takes priority over the preservation defaults below — if they asked to remove or replace a section, do it; only touch what they named.
  3. Use edit_file(path, old_string=..., new_string=...) for targeted insertions or replacements (e.g. adding a subsection under ## Hiring team, updating the salary bullet). Use write_file only when restructuring most of the report.
  4. Preserve every other section, bullet, and source URL the user did not mention. The no-fabrication rule still applies — if a newly-requested fact is not findable, write Unknown — no public signal.
  5. Reply with the update-mode contract: Updated research report at: <output_path>.

Rules

  • Cite URLs inline next to non-trivial claims. End with a ## Sources list.
  • Truth (absolute): if a fact is not findable, write Unknown — no public signal rather than fabricating.
  • By default, keep total under ~600 lines (punchy beats exhaustive) and emit a single file. If the user explicitly asks for a longer report or a different shape, comply.
All versions