feat: add technical job search instructions

[lavkeshdwivedi]

Summary

Adds instructions/technical-job-search.instructions.md — instructions for using Copilot in the end-to-end job search workflow for software engineers.

No job search or career content exists in the repo today. This covers four distinct tasks where Copilot can provide real uplift with the right instructions.

What it covers

• Job description analysis: extracts must-haves vs nice-to-haves, surfaces what the role actually solves, and flags red flags in the JD itself (scope creep signals, unrealistic tech bar)
• CV tailoring: language-matching to the JD, impact-first bullet formatting, seniority signal alignment ("built" vs "designed" vs "drove"), cutting content that doesn't match
• Cover letter writing: enforces a three-question structure (why company / why you / why now) in under 300 words, and explicitly rejects common mistakes like restating the CV in prose
• Offer evaluation: structured comparison across comp (base, equity, bonus, total comp), role clarity, company health signals, and specific red flags (pressure to decide fast, illiquid equity)

Why this adds uplift

The model can write cover letters and CVs without instructions, but defaults to generic structures that don't work well in practice ("I am passionate about..."). These instructions encode specific editorial rules that produce tighter output: matching JD terminology exactly, leading bullets with measurable results, the three-paragraph cover letter format. The offer evaluation section gives a framework the model does not apply by default, including the levels.fyi/Glassdoor/Blind comp benchmarking step and concrete red flags to surface.

Testing

Tested against a real job description for a Senior Software Engineer role. The analysis correctly separated must-haves from nice-to-haves and flagged a "10+ technology" requirement. CV tailoring output matched JD terminology and reformatted bullets to lead with impact.

[github-actions]

⚠️ This PR targets main, but PRs should target staged.

The main branch is auto-published from staged and should not receive direct PRs.
Please close this PR and re-open it against the staged branch.

You can change the base branch using the Edit button at the top of this PR,
or run: gh pr edit 2041 --base staged

[github-actions]

✅ Base branch is now set correctly.

Removing the prior block because this PR no longer targets main.

[github-actions]

⚠️ External plugin PR checks need maintainer follow-up

• Changed entries detected: 1
• Workflow state label: awaiting-review

Per-plugin quality summary

Plugin	skill-validator	install smoke test	overall	source tree
modernize-dotnet	not_run	not_run	infra_error	repository

• modernize-dotnet: skill-validator=not_run, install-smoke=not_run, overall=infra_error

[github-actions]

🔒 PR Risk Scan Results

Scanned 2 changed file(s).

Severity	Count
🔴 High	0
🟠 Medium	0
ℹ️ Info	0

✅ No matching risk patterns were detected in changed files.

This is an automated soft-gate report. Findings indicate review targets and do not block merge by themselves.

[lavkeshdwivedi]

Rebased onto current staged as a clean single-commit branch — just instructions/technical-job-search.instructions.md. The previous branch was stacked 75+ chore: publish from staged commits behind, causing the merge conflict.

[aaronpowell]

What you're trying to achieve here is probably better designed as a skill (or set of skills) than instruction. Instructions are added to every turn of the agent, which would result in this adding additional tokens that might not be providing real value.

Skills on the other hand are opted into by the agent when they hit points that it's needed.