resume-application

# Resume Application Run a controlled, pure-command-line BossKit workflow for one AI Agent opportunity. Treat conversation as qualification, not as permission to submit a resume. Keep every platform write explicit, narrow, and verifiable. ## Boundaries - Use BossKit commands only. Never open a browser or use browser automation. - Restrict the target to an AI Agent role. - Select one best role per company. Reject parallel applications to multiple roles at the same company. - Never expose cookies, tokens, platform remote IDs, or personally identifiable information. Refer to a target by company, title, and local cached job ID. - Never invent an endpoint, raw-call a platform API, or infer success from a local state change. ## Establish Readiness 1. Check local authentication state with: ```bash boss status --platform zhipin ``` 2. Inspect the sanitized online resume snapshot with: ```bash boss account resume show ``` 3. Find and inspect candidates with: ```bash boss search "AI Agent" --platform zhipin boss ls --platform zhipin boss show <LOCAL_JOB_ID> ``` 4. Choose exactly one best role at a company. Record the company, title, local cached job ID, salary evidence, schedule evidence, and relevant skill fit. 5. Stop before contact when the stated monthly salary floor is below 10K, the role is not an AI Agent role, or a duplicate same-company conversation or application already exists. 6. Treat missing or ambiguous salary and double-weekend information as unresolved hard gates, never as implied acceptance. ## Start a Human Conversation 1. Begin with the platform greeting only after the target has passed initial inspection: ```bash boss chat greet <LOCAL_JOB_ID> --yes ``` 2. Never submit a resume on first contact. Never attach a resume to the greeting. 3. Personalize the opening to the exact role and one relevant strength. Avoid batch-identical introductions, generic spam, and any pitch for an automation product or workflow. 4. Read the exact conversation before composing another message: ```bash boss chat history <LOCAL_JOB_ID> --limit 20 boss chat inbox <LOCAL_JOB_ID> ``` 5. Send at most one concrete question per message: ```bash boss chat send --message "<ONE_CONCRETE_QUESTION>" <LOCAL_JOB_ID> --yes ``` 6. Wait for a response before asking the next question or expanding activity. Never scale to more roles while the selected company conversation remains unresolved. 7. Learn the following naturally across the conversation: - whether salary can meet the 10K minimum; - whether double weekends are explicitly supported; - welfare details; - young-team context and collaboration style; - meal and housing support; - development-equipment policy. 8. Ask about the equipment policy neutrally. Never demand or directly ask for a MacBook Pro. 9. Stop after rejection. Send no unsolicited follow-up. Ask once for a rejection reason only when explicitly requested. ## Enforce Submission Gates Require all conditions before considering submission: - Confirm an AI Agent role and one-company/one-role uniqueness. - Confirm salary evidence at or above 10K. - Confirm explicit double-weekend evidence. - Collect sufficient conversational evidence for welfare and working-context assessment, while allowing unanswered non-hard preferences to remain noted. - Confirm no rejection and no prior submission to the exact target. - Refresh the sanitized resume preview with `boss account resume show`. - Inspect current capability with: ```bash boss account resume --help ``` Current BossKit exposes only `boss account resume show`; it has no platform resume-submission command. When no documented `boss account resume submit` or equivalent appears in current help, **STOP** and report the missing capability. Never fabricate the command, call a raw endpoint, or claim that submission occurred. If a supported submission command appears in a future version, first present the exact company, title, local cached job ID, sanitized resume preview, and collected gate evidence. Request explicit confirmation immediately before that single submission. Never auto-submit or reuse earlier consent. Execute only the documented command, then verify the platform write through a documented post-write read command. Report failure or unverifiable state honestly. ## Hand Off Evidence Return a compact record without secrets, remote IDs, or PII: ```text resume_submitted: true | false exact_target: <company> / <title> / <local cached job ID> resume_preview: <sanitized snapshot identity and section counts> user_confirmation: <immediate confirmation present | absent> conversation_evidence: <safe summary of greeting, replies, and hard-gate answers> salary_gate: pass | fail | unknown double_weekend_gate: pass | fail | unknown rejection_status: not_rejected | rejected | unknown no_duplicate_proof: <same-company role and prior-submission checks> submission_verification: verified | failed | unsupported ``` Set `resume_submitted: true` only after a supported platform write and independent post-write verification both succeed. Otherwise keep it `false` and state the exact blocker.