buildresume

name	buildresume
description	Build a complete resume through all 3 steps (draft, provenance check, final)
disable-model-invocation	true

Configuration

Read .jobops/config.json. If missing, stop with:

JOBOPS NOT CONFIGURED Run /jobops:setup to initialize your workspace.

Use config.directories.<key> for all file paths in this skill. Use config.preferences.cultural_profile if this skill generates resume-style content. Use config.preferences.default_jurisdiction if this skill has jurisdiction-sensitive logic (crisis/legal skills accept --jurisdiction=<ISO-3166-2> to override).

Templates

For each template used by this skill, resolve the full path as:

{config.templates.base_dir}/{config.templates.active.<template_name>}/

Templates referenced by this skill: evidence_verification_framework

Application Path Resolution

This skill writes to a per-application folder. Before writing any output:

1. Parse {Company}_{Role}_{YYYYMMDD} from the job-posting filename, or honor --app=<slug> if supplied. The slug MUST be canonical: a leading PascalCase {Company} token (matching the Company_Intelligence/{Company}/ folder so OSINT links), a PascalCase {Role} (underscores between words allowed), and a trailing compact 8-digit date (20260519 — no hyphens, no time). Reject leading date/time prefixes such as 2026-04-15_214414_...; if the source filename carries one, recompose it into canonical form ({Company}_{Role}_{YYYYMMDD}) before composing the folder path.
2. Compose the app folder: {config.directories.applications_root}/{app_slug}/.
3. Resolve this skill's sub-folder by category:
   • resume-development (buildresume, provenance-check) → resume/
   • cover-letter (coverletter) → cover-letter/
   • rubric / assessment (createrubric, assessjob, assesscandidate, auditjobposting) → assessment/
   • briefing / interview prep (briefing, interviewprep) → interview/
4. If the app folder does not exist, mkdir -p it, then copy {config.directories.job_postings}/{filename} → {app_slug}/job_posting.md so the pinned JD cannot silently change under completed work. Ensure the pinned copy begins with YAML front matter carrying output_type: job_posting: if the source JD already has a front-matter block, add the key to it; otherwise wrap a new block (--- / output_type: job_posting / source_jd: {filename} / ---) above the JD body.
5. Exact-slug collisions (same Company+Role+Date) are not auto-suffixed. If the folder already contains the same output type, require the user to pass --app=<distinct-slug>.
6. Output filenames (fixed; only the sub-folder above is resolved dynamically). Pass the fully resolved absolute path to each step's agent in its Task instruction:
   • Step 1 draft → resume/step1_draft.md
   • Step 1 manifest (revise mode only) → resume/step1_manifest.md
   • Step 2 provenance analysis → resume/step2_provenance.md
   • Step 3 final resume → resume/step3_final.md Any PDF/TeX/DOCX derivative shares the exact basename of its source .md and lives in the same sub-folder (e.g. resume/step3_final.pdf) — never a separate latex/ folder.

Arguments

• $1: Job description file path (required)
• $2: Cultural profile (optional). In scratch mode, defaults to "Canadian". In revise mode, the base resume's voice is inherited and the profile menus are skipped; passing $2 explicitly overrides the inherited voice.
• --base=<path>: Force a specific base resume file. Skips fit assessment; implies revise mode.
• --from-scratch: Skip base discovery entirely; run the from-scratch pipeline.

First selects a build mode (see Build Mode Selection below), then runs the three-step build sequentially:

1. Step 1: Revise the selected base toward the JD via change manifest and surgical edits (revise mode) or create an initial tailored draft using HAM-Z methodology (scratch mode)
2. Step 2: Perform comprehensive provenance analysis for credibility (delta-tagged BASE/NEW in revise mode)
3. Step 3: Create final hardened resume addressing all issues (edit-mode with diff gate in revise mode)

Build Mode Selection

Determine the build mode BEFORE dispatching any step agent.

1. Flag shortcuts. --from-scratch → mode = scratch, skip to dispatch. --base=<path> → verify the file exists and is .md; mode = revise with that base, skip to dispatch. (Stamping and fit assessment are skipped; if the forced base lacks role_family, the promotion offer will ask for one.)

2. Library check. Read config.directories.tailored_cv. If the key is absent, or the directory is missing or contains no .md files → mode = scratch, and tell the user once:

   No base resume library found. Building from scratch. To enable revise mode, re-run /jobops:setup and curate base resumes in your Tailored_CV directory — a great first base is the final this run produces (see promotion offer at the end).

   Non-.md files in the directory are ignored with a one-line note.

3. Stamp untagged bases (one-time, per file). For each candidate .md lacking front matter with output_type: resume_base and a role_family value: infer a short category label from the file's content, propose it to the user, let them confirm or type their own label (free text — the taxonomy is entirely user-defined; never assume a built-in category list), then stamp the file's front matter with output_type: resume_base and the confirmed role_family (create the front-matter block if the file has none). Do not alter the resume body.

4. Fit assessment (inline — do not dispatch a sub-agent). Read the JD. For each candidate base, assess fit on four axes:

   • declared role_family match to the JD's role family,
   • positioning level (1–5 IC→C-suite scale, as used by the step1 agents),
   • domain/industry overlap,
   • requirement-keyword coverage. Verdict per candidate: STRONG / PARTIAL / POOR, each with a one-line rationale phrased in the user's own category labels. No numeric scores.

5. Decision gate (the user always confirms).

   • Any STRONG candidate → recommend revising from the strongest one; offer [revise from / build from scratch / pick another base]. "Pick another base" re-presents the remaining candidates with their verdicts; the user's choice proceeds to dispatch as the selected base.
   • Only PARTIAL candidates → present both options, naming the specific gaps (e.g., "base covers the summary and primary role, but the JD's P&L emphasis is not covered"); offer [revise from / build from scratch].

Output metadata

Every markdown file generated during this skill must begin with a YAML front-matter block so downstream tooling can parse metadata without heuristics. Populate the fields with real values before writing any body content. The output path for each step is determined by step 6 of the Application Path Resolution protocol above (step 1 → resume/step1_draft.md, step 2 → resume/step2_provenance.md, step 3 → resume/step3_final.md); job_file in the front matter is always the absolute path to the input JD.

• Step 1 draft — set generated_by: /buildresume step1-resume-draft, output_type: resume_step1, status: draft, version: 1.0.
• Step 2 provenance analysis — set generated_by: /buildresume step2-provenance-check, output_type: resume_provenance, status: analysis.
• Step 3 final resume — set generated_by: /buildresume step3-final-resume, output_type: resume_final, status: final. Increment version if regenerating after revisions.

Common fields for all three:

---
job_file: <absolute path to JD>
role: <target role title>
company: <hiring company>
candidate: <full candidate name>
generated_on: <ISO8601 timestamp>
---

Always write the front matter before any markdown headings or narrative body.

All three files also carry build_mode: revise or build_mode: scratch. In revise mode, step 1 and step 3 outputs additionally carry base_resume: <absolute path to the selected base>. The step agents' own front-matter templates do not list these fields — I pass the build_mode value (and, in revise mode, the base_resume path) in every step agent's Task instruction with the instruction to include them in the output's front matter.

Revise mode writes one additional file — the change manifest:

• Step 1 manifest (revise mode only) — resume/step1_manifest.md, with generated_by: /buildresume step1-resume-revise, output_type: resume_manifest, status: manifest, and base_resume set. Written by the step1-resume-revise agent before it touches the draft.

Step 1: Creating Initial Resume Draft

✓ Initiating strike package assembly

I'll start by reading the job description file and creating the initial draft using the $2 cultural profile.

First, let me read the job description:

@$1

If mode = revise — Deploying Step 1 Agent - Base Revision

I'm launching the step1-resume-revise agent. In its Task instruction I pass, as absolute paths: the selected base resume, the JD ($1), the master inventory root (config.directories.resume_source), the manifest output path ({app_slug}/resume/step1_manifest.md), and the draft output path ({app_slug}/resume/step1_draft.md). I pass $2 only if the user supplied it — otherwise the agent inherits the base's voice and skips the profile menus. This agent will:

• Gap-analyze the base against the JD
• Write an explicit change manifest (every change justified by a named JD requirement)
• Copy the base and apply only the manifest changes as surgical edits
• Run the diff gate proving unchanged content is byte-identical to the base

If mode = scratch — Deploying Step 1 Agent - Initial Draft Creation

I'm launching the step1-resume-draft agent to create an initial tailored resume draft based on the job requirements. This agent will:

• Analyze the job requirements from $1
• Apply the $2 cultural profile preferences
• Use the HAM-Z methodology for strategic positioning
• Review ALL education and credentials from master resume against job requirements
• Explicitly justify any credential exclusions in agent output
• Create a targeted first draft optimized for the role

I pass the resolved absolute output path ({app_slug}/resume/step1_draft.md, per step 6 of Application Path Resolution) to the agent in its Task instruction. I also pass build_mode: scratch for the agent to include in the draft's front matter.

Step 2: Provenance Analysis

✓ Executing credibility verification sweep

✓ Deploying Step 2 Agent - Provenance Verification

Now I'll launch the step2-provenance-check agent to analyze the Step 1 draft for:

• Credibility issues and evidence gaps
• Risk factors for each major claim
• Defensibility of quantified achievements
• Consistency across all experience sections
• Missing credentials: Verify ALL education/designations from master resume are evaluated
• Credential-to-job mapping: Flag any excluded credentials that map to job requirements
• Recommendations for strengthening credibility

I pass the Step 1 draft path ({app_slug}/resume/step1_draft.md) to read and the resolved output path ({app_slug}/resume/step2_provenance.md) to write, both in the agent's Task instruction.

In revise mode I additionally pass build_mode: revise, the base resume path, and the manifest path ({app_slug}/resume/step1_manifest.md) so the agent tags every finding with its origin (BASE = carried from the base, including the user's manual edits; NEW = introduced or modified by the manifest) per its Revise-Mode Delta Tagging section. In both modes I pass the build_mode value for the agent to include in the provenance file's front matter.

Step 3: Final Hardened Resume

✓ Producing deployment-ready final resume

✓ Deploying Step 3 Agent - Hardened Resume Production

Finally, I'll launch the step3-final-resume agent to:

• Incorporate all recommendations from Step 2 analysis
• Address credibility concerns while maintaining competitive positioning
• Produce a fully defensible final resume
• Ensure maximum interview potential

I pass the Step 1 draft ({app_slug}/resume/step1_draft.md) and Step 2 provenance analysis ({app_slug}/resume/step2_provenance.md) paths to read and the resolved output path ({app_slug}/resume/step3_final.md) to write, all in the agent's Task instruction.

In revise mode I additionally pass build_mode: revise and the base resume path so the agent follows its Revise-Mode Edit Protocol: copy the draft to the final, apply targeted edits only for Step 2 findings, and prove with a diff gate that nothing else changed. In both modes I pass the build_mode value (and, in revise mode, the base_resume path) for the agent to include in the final's front matter.

Mission Summary

All three steps of the resume assembly process will be completed:

• Step 1: Initial targeted draft created
• Step 2: Comprehensive provenance verification performed
• Step 3: Final hardened resume produced

Your deployment-ready resume will be ready for mission execution with full credibility and competitive positioning.

Promotion Offer (both modes)

After Step 3 delivers resume/step3_final.md, offer exactly once:

Promote this final to your Tailored_CV library as a base for future applications?

1. Update an existing base (pick which)
2. Save as a new variant (pick a filename and a role_family label)
3. Skip

Skip this offer entirely if config.directories.tailored_cv is not configured.

On update: copy step3_final.md over the chosen base file, then restamp the library copy's front matter: output_type: resume_base, the base's existing role_family (confirm with the user), promoted_from: <app_slug>, and bump version. Remove application-specific keys (job_file, build_mode, status).

On new variant: same restamp, but ask the user for the target filename and the role_family label — offer the labels already present in the library plus "new category" (free text). This is how the user's taxonomy grows.

Never write to the library without this explicit confirmation.