Add technical-writer skill and require it for repo prose

[bensabic]

Adds the technical-writer skill in .agents/skills/technical-writer/ and requires it for prose written in this repo: docs, READMEs, commit messages, PR titles and descriptions, and issues. The same skill was added to vercel/chat in vercel/chat#976.

Coding agents write a growing share of this repo's prose. The skill gives them one voice and editorial standard: it lists the patterns to cut (filler, hype, bold emphasis, em dashes) and asks writers to check technical claims against the source before stating them.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated
deepsec-website	Ready	Preview, v0	Sep 29, 2026 10:54am UTC

1 Skipped Deployment

Project	Deployment	Actions	Updated
deepsec	Ignored	v0	Sep 29, 2026 10:54am UTC

[auto-maintain]

🤖 auto-maintain review

Automated, advisory triage for @bensabic's PR. Facts below are read from the GitHub API.

Check	Result
Author's merged PRs (this repo)	0 — ⚠️ first-time contributor
Account established	✅ (age 3452d · 163 followers · 21 public repos)
Commits signed/verified	✅ 1/1

Review panel: 🟡 medium highest severity

deepsec maintainer code review: 🟡 medium

The skill is added correctly, but its discovery metadata does not cover several prose tasks the PR claims to enforce.

• .agents/skills/technical-writer/SKILL.md:3 — The skill metadata only advertises developer-documentation tasks, so automatic skill matching will not select it for commit messages, PR titles/descriptions, or issues—the prose surfaces this PR intends to require it for. Add those tasks to the frontmatter description; template comments do not help when composing a commit.

General code review: 🟢 low

No actionable defects found in the complete documentation and template diff.

Scanner and matcher quality: 🟢 low

No actionable scanner or matcher quality issues; the changes do not affect detection behavior or pipeline synchronization.

Sandbox and credential boundaries: 🟢 low

Documentation-only change: four new Markdown skill files under .agents/ plus pointer text in CLAUDE.md, CONTRIBUTING.md, and the GitHub templates. All files are mode 100644 (no symlinks or exec bits), no runtime, sandbox, network, credential, or packaging surface is touched, and .agents/ is outside packages/deepsec's published `files` list so the distributed package is unchanged. The skill content is ordinary editorial guidance with no embedded directives that would alter agent trust boundaries. No actionable findings.

Durability and distributed state: 🟢 low

No durability or distributed-state concerns; the diff only adds documentation and templates.

_{Posted by auto-maintain. This automated code review is advisory; a human maintainer makes the call.}

[Anshuman71]

Looks good to me!