hr-reviewer
Skill6 starsChanged 5 months agoMITView on GitHub
Review resumes, cover letters, application packages, and evidence maps from an HR perspective. Use when the user asks for 자기소개서 리뷰, 자소서 검토, HR 검토, 제출 전 점검, unsupported-claim detection, company-name residue checks, or risk-focused revision advice before applying.
In this repository8 other files it gives its agents.
- AGENTS.mdAGENTS.md
- application-packagerSkill
- company-values-analyzerSkill
- cover-letter-writerSkill
- interview-prepSkill
- jd-analyzerSkill
- job-searcherSkill
- resume-intakeSkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. About 0 are common, 9 are distinctive, and 5 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
Use this skill after cover-letter-draft.md and evidence-map.md exist and before cover-letter-final.md is created.
- Distinctive: line 14
Use it when the user asks for HR review, 자기소개서 리뷰, unsupported-claim detection, company-name residue checks, or final submission risk …
- Distinctive: line 18
Do not use this skill as a substitute for resume intake, JD analysis, or cover letter drafting.
- Very specific: line 33
Do not validate a claim as factual unless resume.md supports it.
- Distinctive: line 34
Do not treat company values or JD wording as proof of the applicant's experience.
- Distinctive: line 69
Check whether the answer satisfies the question.
- Distinctive: line 70
Check JD fit and career-level fit.
- Distinctive: line 74
Check fit against the narrowest sub-role the posting names, not just the job family.
- Very specific: line 76
Verify every strong claim against resume.md.
- Distinctive: line 78
Check company name, role name, sub-role name, and copy-paste residue.
- Distinctive: line 79
Check any 실패/성장과정 answer for owned decision, honest cause, and a concrete change afterward.
- Distinctive: line 81
Check readability, sentence length, repetition, and vague claims.
- Very specific: line 98
Check required question IDs, actual answer presence, known per-question counting rules and limits, and that evidence IDs resolve to …
- Very specific: line 170
If resume.md is missing, do not verify claims as factual.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
HR Reviewer
Review as a strict but practical HR screener. Prioritize risks that can cause rejection or weak interview defense.
Trigger
Use this skill after cover-letter-draft.md and evidence-map.md exist and before cover-letter-final.md is created.
Use it when the user asks for HR review, 자기소개서 리뷰, unsupported-claim detection, company-name residue checks, or final submission risk review.
Do Not Trigger
Do not use this skill as a substitute for resume intake, JD analysis, or cover letter drafting. Do not create final text when blockers remain.
Autonomy Level
DoF: LOW
Apply the checklist and blocker criteria deterministically. Do not soften blocker findings to keep the workflow moving.
Permitted inferences:
- Interview defense risk from missing evidence, vague claims, or weak claim-to-evidence mapping.
- Severity level from the blocker and warning criteria below.
Prohibited inferences:
- Do not validate a claim as factual unless
resume.mdsupports it. - Do not treat company values or JD wording as proof of the applicant's experience.
Input Contract
Required context:
- Draft cover letter or resume text.
resume.md.- JD analysis or job posting.
applications/<company-role>/evidence-map.mdwhen reviewing a cover letter package.
Optional context:
- Company values analysis.
- Length limit and counting rule.
Required parameters:
company: target company name.role: target role title.
Outputs produced:
applications/<company-role>/hr-review.md
Inputs
- draft cover letter or resume text
resume.md- JD analysis or job posting
- optional company values analysis
- length limit if available
Workflow
- Check whether the answer satisfies the question.
- Check JD fit and career-level fit. Review against the candidate's career type: a new graduate is read for reasoning and potential, an experienced hire for delivered results and organizational fit. Applying the wrong lens produces the wrong findings.
- Check fit against the narrowest sub-role the posting names, not just the job family. Report where the evidence sits relative to that scope.
- Verify every strong claim against
resume.md. - Flag invented or unsupported achievements.
- Check company name, role name, sub-role name, and copy-paste residue.
- Check any 실패/성장과정 answer for owned decision, honest cause, and a concrete change afterward.
- Check readability, sentence length, repetition, and vague claims.
- Generate 2-3 interview follow-up questions for each key claim.
- Recommend concrete revisions.
Review Completion And Freshness
Review the final answer revision as well as the draft. Record a completed review only after reading the current resume, JD, questions, answer text, and evidence map; bind the result to their current hashes through the workflow state tools. Changing any reviewed input invalidates the corresponding review. Empty review headings and an earlier draft review cannot approve a changed final answer.
Record blockers with explicit statuses. unresolved, not resolved, and
미해결 are unresolved values, never substring matches for resolved. Separate
warnings from blockers, and retain the correction and evidence for each resolved
blocker. User risk acceptance cannot waive a factual error or unresolved blocker.
Check required question IDs, actual answer presence, known per-question counting
rules and limits, and that evidence IDs resolve to resume.md. Code checks
references and hashes; you judge whether each cited experience supports the
claim. If company counting rules are unknown, mark confirmation needed instead
of approving compliance. Run ready validation after completing review; only its
success permits ready-for-user-review.
Blocker Criteria
A finding is a Blocker if any condition below is true:
| Category | Blocker Condition |
| Unsupported quantified claim | A metric, count, percentage, revenue, user count, rank, or award appears in the draft but not in resume.md. |
| Wrong company residue | Another target company, role, or product name remains in the draft. |
| Fabricated tool | A technology, tool, certification, or platform appears in the draft but not in resume.md or the user-provided context. |
| Length violation | Final or draft text exceeds the stated limit when the limit is known. |
| JD mandatory mismatch | The JD states a mandatory requirement and the draft claims fit without matching resume.md evidence. |
| Wrong sub-role residue | The draft names a sub-role, 직무, team, or domain other than the one the posting names. This is the same copy-paste class as wrong company residue. |
| Disqualifying failure disclosure | A 실패/성장과정 answer discloses a trust, safety, ethics, or confidentiality problem, or describes a failure with no resolution or end state. |
| Fabricated employment-history fact | A reason for leaving, employment gap, tenure length, or job-change motive appears in the draft but not in resume.md or user-provided context. Reference checks make these verifiable. |
| Previous-employer criticism | The draft criticizes a former employer, manager, team, or colleague. Screeners read it as how the candidate will later describe them. |
A finding is a Warning when the claim is directionally plausible but thin, generic, too long, weakly connected to the JD, or difficult to defend in an interview.
Also raise a Warning when:
- The evidence is genuinely strong but concentrated in a sub-domain adjacent to the named sub-role, and the draft does not explain the transferable mechanism.
- A failure answer shifts cause onto a teammate, manager, professor, or client.
- A failure answer uses fake-weakness framing such as
너무 완벽주의라서. - An answer reports only an outcome where the question asks how the candidate decided. For new graduates the reasoning is often the evaluated part, so a result-only answer with no visible judgment is a fit risk, not a style issue.
- An experienced candidate's answer presents a previous employer's process as the standard rather than as context, which reads as unwillingness to adapt.
- An experienced candidate's answer lists responsibilities where an achievement is expected, or states a result without the mechanism behind it. The mechanism is what transfers; the number does not.
- An experienced candidate changes industry or sub-domain and the draft claims adaptability without naming the transferable mechanism.
- The answer is written in the wrong career-type frame: a new graduate implying ownership of organizational outcomes, or an experienced candidate leaning on potential and attitude where delivery is expected.
Output
## HR Review
- Overall Score:
- Strong Points:
- Weak Points:
- Risky Claims:
- Missing Evidence:
- JD Fit:
- Company Fit:
- Interview Defense Risk:
- Blockers:
- Warnings:
- Interview Defense Questions:
- Recommended Revision:
- Final Submission Checklist:Severity
Blocker: likely false, unsupported, wrong company, or submission-risk issue.Major: weak evidence, poor fit, unclear role, or question mismatch.Minor: style, compression, wording, or readability issue.
Fallback
If resume.md is missing, do not verify claims as factual. State that evidence validation is incomplete and ask for resume.md.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

