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 mostafa-drz/claude-skills --skill step-throughgit clone --depth 1 https://github.com/mostafa-drz/claude-skillsWrote 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/mostafa-drz/claude-skills/step-through)<a href="https://agentmods.dev/skills/mostafa-drz/claude-skills/step-through"><img src="https://agentmods.dev/badge/skills/mostafa-drz/claude-skills/step-through/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/mostafa-drz/claude-skills/step-through"><img src="https://agentmods.dev/badge/skills/mostafa-drz/claude-skills/step-through.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 | $0.00149 | $0.03993 |
| Opus 5.5 | $0.00060 | $0.01597 |
| Sonnet 5.5 | $0.00030 | $0.00799 |
| Haiku 4.5 | $0.00015 | $0.00399 |
Grade A, and why
step-through scanned grade A with 1 finding 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 1mo 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.
Strips warnings and disclaimerslowAnti-refusal
Omitting safety caveats hides risk from the user and is a common jailbreak preamble.
7. **Learn quietly.** Promote a rule only after 3+ consistent feedback signals. Mention once when promoting, never lecture. Demote unstable rules instead of accumulating clutter. Downgraded: this mod is about security review, or the phrase is quoted, so it is likely naming the pattern rather than instructing it.
How it starts
The opening of the file, as written. The whole thing — 339 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/step-through
Take a long enumerated response and walk through it interactively, one item at a time, with a decision per item — and learn the user's per-severity habits over sessions.
Preferences
On startup, use the Read tool to load ~/.claude/skills/step-through/preferences.md. If missing, treat as first-run (see First-time detection below).
Defaults when no preferences exist:
- default-depth:
concise— one paragraph plus 1–2 references on dive deeper - show-outline:
yes— print the outline before starting - persist-session-log:
yes— append decisions tosessions/<date>.md - auto-pick-source:
most-recent— most recent enumerated assistant message - action-confirmation:
destructive-only— confirm before file deletes / branch ops / pushes / external posts; everything else proceeds - resume-on-stop:
yes— when a session is stopped mid-way, save resume state so the next invocation can pick up where it left off - per-severity-defaults: none yet (filled by Learned over time)
Context
Do NOT pre-load anything on startup. The skill works against the current conversation's recent assistant messages. Run Bash/Read only when the user picks "dive deeper" or "take action" on a specific item — at that point, fetch only what's needed for that item.
Command routing
Check $ARGUMENTS:
help→ display help, then stopconfig→ interactive setup, then stopreset→ delete~/.claude/skills/step-through/preferences.mdANDfeedback-journal.mdANDsessions/, confirm, stopfeedback→ collect thumbs-up/down on the most recent walkthrough (see Feedback & learning), stopresume→ continue the most recent stopped walkthrough using saved resume state, stop if none- anything else (including empty) → run the skill
Help
/step-through — Walk through a long enumerated response item-by-item
Usage:
/step-through Walk through the most recent enumerated response
/step-through --filter blockers Only walk through items in that section/severity
/step-through --start 3 Resume from item N
/step-through --from "<hint>" Pick a different message by short description
/step-through resume Continue the most recent stopped walkthrough
/step-through feedback Rate the last walkthrough (improves future defaults)
/step-through config Set preferences
/step-through reset Clear preferences + feedback journal + sessions
/step-through help This help
Per-item options:
Dive deeper Pull code / docs / context to better understand
Take action Edit, run, research, or fix right now
Note & defer Capture decision (e.g. "address with AIS-XXXX"), move on
Skip Drop without action
Stop End the walkthrough, summarise
Examples:
/step-through After a long PR review
/step-through --filter blockers Only critical items
/step-through --start 7 Resume mid-list
Current preferences:
(loaded from ~/.claude/skills/step-through/preferences.md)
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.
- 1mo ago Changed · +3 lines · +149 tokens per session scan B → A 411f204b8e28
- 1mo ago First seen · 336 lines · 0 tokens per session scan B 77333bd74c90
step-through is a skill published in the GitHub repository mostafa-drz/claude-skills (4 stars, last pushed yesterday), licensed MIT. It adds 149 tokens to every session and 3,993 once invoked, about $0.0006 per session on Opus 5.5. A static security scan graded it A with 1 finding (strips warnings and disclaimers). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
agency-multi-brand-pack
Per-client asset templates scoped by workspace.
hydra
Fan multiple independent tasks out to SEPARATE Terminal.app windows, each running its own live claude session with exactly the context it needs — real parallel sessions with isolated context windows, not subagents. Confirms grouping/model/risk with the user before launching, delivers context invisibly, auto-tiles…
application-pack
Turns one pasted job post into an application folder built only from the student's own career vault: a tailored resume, a cover letter under 250 words, and an HR call sheet with the answers already drafted. Reads career-vault/01 to 06 first and refuses to run while the positioning, portfolio story or resume are still…
shipkit-work-memory
Log session progress and save resume state. Infers from conversation and git. Triggers: 'log progress', 'session summary', 'checkpoint', 'save progress', 'end session'.
session-update
Use when the user says "update session", "log progress", "checkpoint session", or invokes /session-update — especially before /clear or handoff.
sounding-board
Use when a user needs to vent, talk something out, or just not feel like they're shouting into a void. Active listening with gentle reflection — not therapy, not silence, just a friend who pays attention.