career-engine-new-application

New Application — Per-Role Pipeline

Registry: this pipeline is listed in the Pipeline Registry in skills/career-engine/SKILL.md. Actions owned by another pipeline's registry row are out of scope here — route to that pipeline instead of improvising.

This skill covers Step 0.10 and Steps 1 through 7 of the New Applications pipeline. Step 0.10 runs once before the per-role loop begins. Steps 1 through 7 repeat for each role in the processing queue. The structured JD was fetched in Step 0.5 and is in memory — pass it directly without re-fetching.

The pipeline produces two deliverables per role: a CV DOCX and a cover letter DOCX. The CV goes through: draft, gatekeeper, recruiter review, revision, gatekeeper (post-revision). The cover letter is written after the final CV, from exactly three content inputs — the final CV, Role emphasis, and Why I Want This Role (see Step 5's input contract). The recruiter review, including its interview-trigger gaps, is user-facing feedback surfaced in Step 7d's feedback file — it is never passed to the letter-writer.

career-data data root (R-37). The personal-data files — 01-writing-rules.md, 02-professional-background.md, 03-framework.md, linkedin-profile.md, pipeline-preferences.json, delivered-letters/, and the user's .dotx — load from ${CAREER_DATA}/references/, the path the orchestrator resolves in its career-data discovery preflight. Every other file (self-checks, REFERENCES.md, skill docs, default .dotx templates) stays on ${CLAUDE_PLUGIN_ROOT}. If ${CAREER_DATA} is not set (direct or standalone invocation outside the orchestrator), locate the career-data skill yourself, confirm career-data-marker.json, and apply the orchestrator's healthy / damaged / absent outcomes before reading. A configured user's missing career-data is a hard stop — never silently fall back to blank templates.

Universal spawn parameter — gap_handling_mode (2026-07-14). $GAP_HANDLING_MODE is resolved once per run (orchestrator config resolution: pipeline-preferences.json → gap_handling; disabled when the value is disabled or empty, enabled otherwise — an absent key on an older config means enabled). Every subagent spawn in this pipeline passes gap_handling_mode=$GAP_HANDLING_MODE, exactly like CAREER_DATA — every agent, every round, every re-spawn and crash-recovery fallback, no exceptions. When disabled, no agent may produce, request, or enforce gap framing anywhere: the coach's outline and review carry no gap directives, the recruiter asks no writer to address a gap in a document, the gatekeeper's Gate 0 gap-handling keyword exception never applies, and no document acknowledges a gap (the letter-side never-name-the-gap rule applies regardless of mode). This is a one-word routing token — for the letter-writer it rides inside its input contract's routing-token allowance, never as content.