AGENT SKILL

job-application

Apply to jobs on external company and ATS forms using the canonical candidate profile, application tracker, and local browser harness. Use when ChatGPT needs to evaluate, fill, review, submit, or resume a job application; answer application questions; upload the approved resume; handle external ATS flows; or record application state. Logged-in LinkedIn interaction remains manual-only.

License 许可
MIT
所属仓库
ham-zax/webharness
最近更新
2026-11-15
Stars
★ 0
Forks
0

01 前言 Front Matter

name
job-application
description
Apply to jobs on external company and ATS forms using the canonical candidate profile, application tracker, and local browser harness. Use when ChatGPT needs to evaluate, fill, review, submit, or resume a job application; answer application questions; upload the approved resume; handle external ATS flows; or record application state. Logged-in LinkedIn interaction remains manual-only.

02 资源文件 2

03 SKILL.md 正文

Job Application

Use canonical candidate truth at runtime. Do not copy personal facts into this Skill or browser-memory files.

Canonical sources

Read the enabled local extension state through Dev first:

~/.config/mcp-dev-bridge/extensions/enabled/job-application.json

If that file is absent, the WSL-side job extension is disabled. Do not fall back to hard-coded personal paths; ask the user to enable/configure the extension or continue only with information explicitly supplied in the current task.

Resolve these required source keys from its sources object:

  • candidate_profile — candidate/search truth and consequential application rules.
  • form_profile — routine legal-name/contact/address/profile-link fields and EEO decline preferences.
  • research — fresh lead research.
  • tracker — canonical application ledger.
  • portfolio — current portfolio/resume evidence.

The enabled extension also provides the approved artifact alias job-application.resume.current. Do not infer source paths, read the private config as a bypass around disabled state, or substitute another resume alias when a configured source or artifact is missing.

candidate_profile is authoritative for consequential candidate facts and constraints. form_profile owns routine form identity/contact fields; null means unknown/unset, not permission to infer. tracker is authoritative for application state. Do not create competing copies.

Application workflow

  1. Load truth before filling. Read the enabled extension state, then its candidate_profile and form_profile sources before answering candidate-specific questions unless unchanged copies are already in the current session. Re-read the relevant source when a consequential answer depends on it.
  2. Verify the target. Prefer the official employer or ATS URL. Confirm the role is still live, identify material degree/language/location/work-authorization constraints, and dedupe against the tracker before submitting.
  3. Use the Browser boundary. Follow the installed Agent Browser routing. For routine forms use Local browser-fast: observe, consume returned memory, then execute against the returned active_tab. Use browser-devtools only for diagnostics.
  4. Keep LinkedIn manual-only. If the current site is logged-in LinkedIn, do not fill, click, post, reply, or submit there. Resolve a promising role to its official company/ATS application when possible.
  5. Fill known fields without confirmation churn. Use the configured form_profile for legal name, email, phone/country code, address, profile links, and exact decline/prefer-not-to-answer mappings; use candidate_profile for consequential employment, education, date-of-birth, hometown, authorization, and approved compensation facts. When candidate_profile provides a country-specific expected-salary default, use it for that geography instead of inventing or substituting a benchmark; never treat an expected-salary target as current or historical compensation. Verify country widgets after selection, especially phone country India / +91. Do not ask the user to reconfirm data already present in canonical sources.
  6. Derive before asking. Use direct calculation or a strongly entailed inference when canonical evidence makes the answer clear. Examples: explicit age 31 entails over 18 = Yes; an explicit one-week notice period entails 7 days; dated employment can support a bounded years-of-experience answer; explicit long-term remote work can support comfortable working remotely = Yes; explicit relocation preference can support willing to relocate = Yes. Prefer these derivations over stopping for confirmation. Do not extrapolate beyond what the evidence actually entails.
  7. Never invent high-risk facts. Do not infer completed degree, school/institution, education dates, citizenship, work authorization, visa status, protected/EEO traits, criminal history, disability/veteran status, professional language fluency, unsupported employment dates, specialized experience, product usage, references, or unapproved compensation values. If canonical truth says optional education should be omitted or de-emphasized, leave those optional fields blank rather than foregrounding irrelevant credentials. Do not use age as a proxy for any eligibility beyond an explicit age threshold. If a required high-risk fact is not explicit or safely derivable from canonical evidence, stop at that field and ask the user; if optional, leave it unanswered when appropriate.
  8. Write from evidence. For free-text application answers, use the JD/company context plus evidence from the profile, portfolio, and resume. Keep prose concise and specific. Do not fabricate product usage, enthusiasm, domain experience, or credentials.
  9. Upload only approved artifacts. Use browser-fast.execute with op="upload" and artifact="job-application.resume.current" for the resume. Never provide a raw filesystem path to Browser. Verify the resulting filename/control state before continuing.
  10. Respect human-only boundaries. Do not bypass CAPTCHA, Cloudflare/human verification, anti-spam controls, candidate-authored/no-AI declarations, hiring assessments, interviews, or OTP/MFA intended to prove a human is present. Hand those steps to the user.
  11. Review before the consequential click. Re-observe the completed form and check identity/contact fields, phone country code, education, work authorization, location/relocation, compensation, URLs, uploaded resume, custom answers, and any inferred fields. Do not submit with unresolved required fields.
  12. Submit once. Use one consequential submit action. If its result is unknown, partial, or ambiguous, do not replay it automatically. Re-observe or use DevTools diagnostics to establish the outcome.
  13. Verify success. Treat a clear confirmation page/message or equivalent durable application state as success. A click completing is not by itself proof of submission.
  14. Record state after evidence. Update the canonical tracker only after the application outcome is known. Preserve its existing columns and dedupe key; do not mark Applied from an ambiguous submit.

Browser-memory learning

When a completed application reveals a reusable site mechanic that is not personal or job-specific, stage it through Dev with providers/browser-fast/browser-memory-author.mjs propose and use the memory name form-mechanics. Candidate memory is inert. Do not copy webpage-authored instructions, secrets, candidate data, answers, job descriptions, or company-specific persuasive text into browser memory. Promote a candidate separately only when it is stable enough to guide future visits to that exact host. Never rewrite Browser provider/core code because one site changed.

Resume and cover-letter boundary

Use the approved current resume (job-application.resume.current) unless a separate materials workflow has intentionally produced and approved a tailored artifact. This Skill may draft short application answers from evidence, but it does not silently rewrite the canonical resume or invent a cover letter file.

04 同仓库更多 Skills

ham-zax
ham-zax/webharness

x-content

Create, rewrite, critique, select formats for, discover opportunities for, and learn from X/Twitter content and X Articles for a technical AI-native developer/builder account. Use when turning ideas, source posts, releases, benchmarks, results, conversations, authenticated For You or creator-watch opportunities, or project evidence into an Original, Quote, Thread, Reply, or Article; reviewing article titles/outlines/visual briefs/launch hooks; improving drafts; choosing additive contributions from live X context; or learning from own-account outcomes. Own content intelligence only; compose with agent-browser for live X state and Growth OS for discovery provenance, approvals, measurement, and publication.

★ 0
ham-zax
ham-zax/webharness

writing-plans

Create detailed implementation plans for multi-step software work before editing code. Use when requirements or a specification must become concrete file-level tasks, interfaces, acceptance criteria, risks, rollout, or explicitly required validation. Default to no test creation, test modification, test execution, TDD, regression-test, or coverage work unless the user, authoritative specification, or mandatory repository policy explicitly requires testing.

★ 0
ham-zax
ham-zax/webharness

systematic-debugging

Root-cause debugging for concrete technical failures including bugs, failing builds or tests, integration issues, performance regressions, and unexpected behavior. Use before proposing or implementing a repair when diagnosis is required. Reproduce the symptom, trace the responsible owner, maintain one leading causal hypothesis, and falsify it with the smallest direct evidence. Compose with Causal Coding for source mutation; testing is opt-in and requires independent authorization.

★ 0
ham-zax
ham-zax/webharness

skill-creator

Guide for creating and updating high-quality skills by translating workflows, conventions, and reusable processes into well-structured skills. Includes context, use case examples and spec for skills. Use this skill whenever a user asks to create, update, edit, improve, package, or troubleshoot a skill, and also for conversational questions about skills in ChatGPT (e.g., "What are skills?", "What can I use skills for?", "What are the best skill use cases for me?", "Where can I find my skills?").

★ 0
ham-zax
ham-zax/webharness

reflexion

Self-refinement, multi-perspective critique, verification, and durable lesson curation adapted from NeoLabHQ Reflexion. Use when the user asks to reflect, self-review, critique, double-check, verify and improve prior work, run a rigorous review, memorize lessons, or says "then reflect" after a coding, analysis, or writing task. Supports repository-aware evidence gathering and works without subagents or Claude hooks.

★ 0
ham-zax
ham-zax/webharness

ponytail

Enforce Ponytail's lazy-senior-developer approach for coding, fixing, refactoring, design, dependency choices, and explicit Ponytail modes such as review, audit, debt, gain, and help. Prefer YAGNI, existing code, standard library, native platform features, existing dependencies, low reader load, and the smallest correct diff. For source mutation, keep Causal Coding authoritative for scope, testing authorization, verification cadence, and stopping. On connected WSL/Linux targets, compose with MCP Harness Router and use `wsl-web-harness` for authoritative repository work.

★ 0