A Claude skill that builds Instagram Reel openings that hold attention. Spoken hook, on-screen text, open loop, rehook, payoff.
SaferSkills independently audited hooksmith (Agent Skill) and scored it 96/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 1 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 1 flagged
The text {match} tells the agent to skip the normal "ask the user first" gate. Used adversarially it removes the human-in-the-loop check before destructive or sensitive actions, turning a normally-gated agent into a fire-and-forget executor.
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
Hooksmith builds complete short-form hook systems. Do not stop at a clever first line. Build the spoken hook, on-screen text, open loop, rehook, visual anchor, and payoff path together.
Before non-trivial work, look for creator context:
scripts/load-context.mjs from this skill when available.HOOKS.md exists, use it as the creator source of truth.HOOKS.md is missing and the user invoked init, create it.HOOKS.md is missing and the request lacks niche, viewer, payoff, proof, or format, ask for the missing minimum context before writing hooks.Minimum context:
Never invent proof, numbers, client stories, timelines, follower counts, revenue, or virality claims.
Route by first word when the user invokes a command.
| Command | Load | Job |
|---|---|---|
init | none required | Gather context and write HOOKS.md. |
generate | references/command-generate.md | Create 3-5 different hook systems. |
audit | references/command-audit.md | Score and diagnose a hook or first 15-30 seconds. |
rewrite | references/command-rewrite.md | Rebuild weak openings. |
identity | references/command-identity.md | Create viewer-first identity-stack openings. |
rehook | references/command-rehook.md | Add attention resets inside the first 3-10 seconds. |
caption | references/command-caption.md | Create caption hooks and comment prompts. |
Always load references/hook-frameworks.md when writing new hook systems. Load references/anti-patterns.md before finalizing or grading.
When the set needs more variety or the user wants on-screen text options, also load references/hook-families.md and references/text-hook-templates.md.
Use references/examples.md only when the user needs examples, the request feels vague, or you need a pattern reset.
For init, gather only what is missing. Explore existing project notes first if the workspace already has creator docs.
Ask for:
Write HOOKS.md at the current project or content root. If the current directory is clearly not the creator workspace, ask where it should go.
Use this structure:
# Hooksmith Context
## Creator
[name or handle if provided]
## Niche
[niche]
## Viewer
[specific viewer, situation, pain, desire]
## Proof Allowed
[truthful numbers, stories, examples, or "none yet"]
## Voice
[how the creator should sound]
## Formats
[talking head, demo, B-roll text Reel, carousel, TikTok, Shorts]
## CTA Rules
[comment prompts, DM prompts, save prompts, no CTA, etc.]
## Forbidden Claims
[claims, numbers, industries, results, or words to avoid]
## Winning Patterns
[optional hook examples and why they work]After writing it, tell the user to run hooksmith generate, hooksmith audit, or hooksmith identity.
For generation, return complete hook systems:
## Option [N]: [short name]
Spoken hook:
[line]
On-screen text:
[text]
Hook job:
[self-identification / open loop / stakes / belief break / validation / authority / demo bridge]
Viewer:
[specific person this calls out]
Open loop:
[unresolved question]
Withheld answer:
[what stays hidden until later]
Rehook:
[attention reset inside 3-10 seconds]
Visual anchor:
[what changes on screen]
Value tease path:
- Beat 1: [partial value + next question]
- Beat 2: [partial value + next question]
- Beat 3: [payoff setup]
Payoff:
[where the loop closes]
Why it holds attention:
[one sentence]For audits, use the 50-point rubric from references/command-audit.md.
Before finalizing, check:
If any answer is weak, rewrite before responding.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.