Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add melodic-software/claude-code-plugins --skill hold-my-handgit clone --depth 1 https://github.com/melodic-software/claude-code-pluginsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/melodic-software/claude-code-plugins/hold-my-hand)<a href="https://agentmods.dev/skills/melodic-software/claude-code-plugins/hold-my-hand"><img src="https://agentmods.dev/badge/skills/melodic-software/claude-code-plugins/hold-my-hand/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/melodic-software/claude-code-plugins/hold-my-hand"><img src="https://agentmods.dev/badge/skills/melodic-software/claude-code-plugins/hold-my-hand.svg" alt="Reviewed on agentmods" width="80" height="20"></a>What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00130 | $0.01947 |
| Opus 5.5 | $0.00052 | $0.00779 |
| Sonnet 5.5 | $0.00026 | $0.00389 |
| Haiku 4.5 | $0.00013 | $0.00195 |
Grade A, and why
hold-my-hand scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 168 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Hold my hand
Once invoked, this is a standing instruction for the rest of the session. Take the work that remains, cut it into phases, and present one phase at a time. The user never sees Phase 2 until Phase 1 is presented and done.
A declared species in this plugin: NOT a corrector
This skill is a standing session posture over how future work is
delivered, not a re-anchor/audit/correct loop. It audits nothing; it changes
the shape of every subsequent response until the session ends or the user
cancels it. It carries no discipline-batch tier, on purpose: a batch sweep
must never impose a delivery posture the user did not ask for.
Partition the work into phases
A phase is a contiguous run of steps that ends at a point the user can confirm. Cut on the confirmable boundary, not on an even step count: a phase ends where something is observably true (a file exists, a service answers, a page loads).
State the count once, on a single line, so the user knows the shape of the job without reading ahead:
Phase 3 of 8: Install the runner on the build machine
Never list the later phases. Not as a preview, not as a table of contents, not as a "here is where we are going" paragraph, and not as a task-ledger dump the user reads over your shoulder. The count is the only forward information.
When the work turns out to need more phases than the first estimate, say so in one line ("this is now 10 phases, not 8") and continue. Never grow the count silently.
What every phase states
Each phase carries all six of these. A phase missing one of them is not ready to present.
- Where. The machine, the kind of terminal (PowerShell, bash, a browser tab, a physical console), whether it must be elevated, and the working directory.
- Numbered steps. One action per numbered step, in the order they run.
- The exact command, in a fenced block, with real values substituted, never a template with placeholders the user must fill in from memory. A step that has no command because it happens in a browser, a vendor dashboard, or on a physical console names the exact control to operate and where to find it, in the command's place. Never invent a command to satisfy this part.
- The expected output or observable result after each command or action, so the user can tell success from failure without asking.
- The common failure and its repair. The one that actually happens, with what to do about it.
- The exact reply that advances. Write it as a literal the user can copy:
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago Changed · -6 lines 1c5843d923d7
- 28d ago Changed · -87 tokens per session 35e6a391a266
- 1mo ago First seen · 174 lines · 217 tokens per session scan A de2032bf06bf
hold-my-hand is a skill published in the GitHub repository melodic-software/claude-code-plugins (28 stars, last pushed today), licensed MIT. It adds 130 tokens to every session and 1,947 once invoked, about $0.0005 per session on Opus 5.5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-09-07.
Other skills, from other repositories
linkedin-engine
Daily interactive LinkedIn session: feed posts, DM outreach, prospect research, engagement tracking, and performance review. Human-in-the-loop. Designed for daily /loop.
reddit-reply
Daily follow-up on Reddit engagement. Reads recent reddit-monitor reports, checks each URL for replies to our comments, and drafts follow-up responses. Designed for daily /loop. Load identity and brand-voice first.
social-media-guide
Entry point for social-media — routes to LinkedIn, Reddit, X/Twitter, and the daily social brief.
catch-expected-errors
For iteration with errors: catch exceptions during exploration, skip invalid cases, continue to next attempt.
verify-with-inline-tests
For test co-location: tests in same file as code, run with main, tests as documentation and examples.
slack
Use when you need to control Slack from OpenClaw via the slack tool, including reacting to messages or pinning/unpinning items in Slack channels or DMs.