Skip to content
Back to skills

Pf Harden

ASecurity

PM step after a deploy: makes the live product fast, findable and able to take a traffic spike — safety-net tests, performance budget on the live site, SEO + AI-engine visibility (Google, Bing, ChatGPT, Claude, Perplexity), a defensive security review (RLS, auth, input, headers, secret and dependency scan — own site only), database indexes, free-layer caching, a local k6 load test and a hosting-limits check — by writing HARDEN-REQUEST.md, driving Claude Code through Chrome to run /pf-next har...

  • 2 stars
  • 0 votes
  • 0 copies
  • 0 views
  • Added November 18, 2026
ai-agentsgogitdatabasesecurityperformance

Works with

  • claude code

Security analysis

A100/100

Scanned November 18, 2026

npx -y skills add singhviplove234/atelier --skill pf-harden --agent claude-code

Installs into .claude/skills of the current project.

Are you the author of Pf Harden?

Add the live security badge to your README. It updates with every re-scan.

Security grade badge for Pf Harden
[![Security: A — Skills Directory](https://www.skillsdirectory.com/api/skills/singhviplove234-pf-harden/badge)](https://www.skillsdirectory.com/skills/singhviplove234-pf-harden)

More formats (shields.io, HTML) on the badges page. Keep it an A: scan every change in CI with Pro.

Download with Pro
SKILL.md
---
name: pf-harden
description: "PM step after a deploy: makes the live product fast, findable and able to take a traffic spike — safety-net tests, performance budget on the live site, SEO + AI-engine visibility (Google, Bing, ChatGPT, Claude, Perplexity), a defensive security review (RLS, auth, input, headers, secret and dependency scan — own site only), database indexes, free-layer caching, a local k6 load test and a hosting-limits check — by writing HARDEN-REQUEST.md, driving Claude Code through Chrome to run /pf-next harden, polling for HARDEN-RESULT.md, then offering merge + redeploy. Use when the user says /pf-harden, 'harden the site', 'make it faster', 'will it survive traffic', 'survive going viral', 'fix the SEO', 'make ChatGPT find us', 'is it secure', or 'lock it down'. Supports --dry-run and --passes."
---

# /pf-harden [--passes safety,perf,seo,security,compliance,indexes,caching,load,limits] [--dry-run]

Post-launch hardening. Read first: `${CLAUDE_PLUGIN_ROOT}/references/performance-budget.md`, `seo-ai-visibility.md`, `security.md`, `deploy-rule.md` (cost posture), `bridge.md`, `handoff.md`.

## 0. Flags first
`--dry-run` → print the `HARDEN-REQUEST.md` that would be written, write nothing, ask nothing, stop. Seat identity = `--lane` (default `solo`); must equal `factory.json.pm_lane`, else "Only the PM lane (`<pm_lane>`) hardens." and stop. `--passes` defaults to all nine; `safety` is always included even if omitted.

## 1. Gate
Locate the project (named, else the only one, else ask). Read `factory.json`. Require `gates.build == "approved"`. Prefer a live site: use `deploy.live_url` when `deploy.status == ok`; otherwise say "No live deploy yet — hardening will measure a local copy instead; results are indicative" and continue with `live_url` empty. Never run on a project whose last build failed.

## 2. Explain and ask — always
Show in one short block what will happen, in the user's words: tests first so nothing breaks; speed check against the budget (score >= 90, loads under 2.5 s, no lag, no jumping); SEO and AI-visibility check (Google + Bing + ChatGPT, Claude, Perplexity crawlers can read every page, structured data, llms.txt, FAQ blocks); a defensive security review of your own site only (data-access rules, logins, input checks, security headers, secret and dependency scan) — read-only, no attack tools, no other sites; a legal/compliance check (dependency licences, privacy/terms/consent, data-deletion) that flags anything a lawyer should see — not legal advice; database indexes; free caching only (no Redis); a fake-traffic test on a local copy only, never the live site; a list of what would break first at 10x traffic with prices, **recommend only — nothing paid gets created**. Then AskUserQuestion: `Harden <slug> now? (passes: <list>)` Yes / No. Only Yes proceeds.

## 3. Hand-off (file-based — see `bridge.md`)
`request_id = <slug>-harden-<yyyymmddHHMM>`. Write `<folder>/HARDEN-REQUEST.md`:
```
PF-HARDEN v1
request_id: <id>
project: <slug>
code_path: <code root>/<slug>
live_url: <deploy.live_url or empty>
target: <deploy.target>
data_sensitivity: <brief.data_sensitivity>
deep_audit: <yes on a pre-go-live/final harden or if the user asked for a deep security audit, else no>
passes: <list>
```
Delete any stale `HARDEN-RESULT.md`. Store `harden.request_id`, `harden.status = requested`; log `{stage: "harden", event: "request written"}`. Hand off per `handoff.md`. **Mode A (event, default):** drive Chrome per `bridge.md` on `lanes[<pm_lane>].code_url` with `/pf-next <slug> harden --go` (fallback: "In your Claude Code session (lane `<pm_lane>`) type **`/pf-next <slug> harden`** and confirm."), then wake on the GitHub PR event; fall back to a slow poll (~every 15 min) for `HARDEN-RESULT.md` (first line exactly `PF-HARDEN-RESULT v1`, same `request_id`) up to 120 minutes. **Mode B (cloud):** run the harden passes inline in the cloud per `handoff.md` and read the result directly (load tests still run against a local copy with a test database, never the live URL). `--handoff event|cloud` overrides for this run.

## 4. Present
Record `harden.status`, `harden.pr_url`, `harden.branch`; log `{stage: "harden", event: "<status>: budget_met <yes/no>"}`. Show: the before/after speed table (route | score | LCP | INP | CLS | JS KB), the SEO table (route | title | description | schema | SEO score | crawlers 4/4), the security table (check | result | fixed/needs-you), indexes added, caching applied, load summary in plain words, the limits list with monthly costs, anything reverted, and `needs_decision` items as a separate "Paid — your call" list. Say plainly whether the site now meets the budget.

## 5. Merge and redeploy (ask)
If `status: ok`: AskUserQuestion `Merge the hardening PR into main and redeploy? Yes / No`. On Yes: write `MERGE-REQUEST-harden.md` per `bridge.md` (`repo`, `pr_url`, `strategy: squash`), drive `/pf-next <slug> merge --lane harden --go`, poll `MERGE-RESULT-harden.md` (20 min), then run the `/pf-deploy` flow from its step 4 (it asks "Deploy now?" itself). On No: "PR is open at <pr_url>; run /pf-harden again or merge later." If `status: failed`: show `notes`/`failures` and offer "run again with fewer passes" or "tell me what to fix".

Never: load-test a live URL, add Redis or any paid resource, upgrade a plan, push to `main` directly, or run without the Yes in step 2.

Attribution

Is this your skill, or is something wrong with this listing? Request removal or report an issue. Author removals are honored within 72 hours.

View all in ai-agents →

Comments

Loading comments…

Newsletter

Get the weekly list of safe new skills, and the ones we pulled.