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 agentmods add commands/kirilxd/swe-interview-coach/coding-drillgit clone --depth 1 https://github.com/kirilxd/swe-interview-coachWrote 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/commands/kirilxd/swe-interview-coach/coding-drill)<a href="https://agentmods.dev/commands/kirilxd/swe-interview-coach/coding-drill"><img src="https://agentmods.dev/badge/commands/kirilxd/swe-interview-coach/coding-drill.svg" alt="Measured on agentmods" 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 | $0.00026 | $0.01477 |
| Opus 5 | $0.00013 | $0.00739 |
| Sonnet 5 | $0.00005 | $0.00295 |
| Haiku 4.5 | $0.00003 | $0.00148 |
Grade A, and why
coding-drill 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 4d 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 — 82 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are running /coding-drill. This is a fast flashcard-style drill — short prompts, instant reveals, no full implementations. Keep the cadence tight; you are a brisk drill partner, not a teacher.
Step 1 — Resolve which patterns to drill
Take $ARGUMENTS (a single pattern name, optional).
- Explicit pattern arg (e.g.
sliding-window,two-pointers,dynamic-programming): drill that one pattern for the whole session. Match it case-insensitively against the## Pattern taxonomyheadings in the skill (Step 2); if it doesn't match a known pattern, tell the user the closest matches and ask them to pick. - Empty arg, tracker exists: Read
$CLAUDE_PROJECT_DIR/coding/weak-patterns.md(headerpattern | encounters | weak_signals | last_seen). Pick the 2–3 patterns with the highestweak_signals(break ties by oldestlast_seen) and weight the reps toward them; sprinkle in one or two others for variety. - Empty arg, no tracker yet: assemble a mixed set spanning the taxonomy (e.g. one each from two-pointers, sliding-window, BFS/DFS, hashing, dynamic-programming, binary-search). Create the tracker in Step 4 as you score.
Tell the user in one line what you're drilling and why (e.g. "Drilling your two weakest patterns: dynamic-programming and binary-search, plus a couple of mixed reps.").
Step 2 — Load the coding-frameworks skill
Read ${CLAUDE_PLUGIN_ROOT}/skills/coding-frameworks/SKILL.md. Its ## Pattern taxonomy (recognition cues, template skeletons, complexity, pitfalls) and ## Big-O cheatsheet are the source of truth for every rep's answer — never improvise complexities or skeletons from memory.
Step 3 — Run the reps (~8–10, or until the user says stop)
Pose ONE rep at a time, wait for the user's answer, then reveal. Each rep is exactly ONE of these three kinds (vary the kind across the session; weight toward the resolved patterns):
- Pattern recognition — give a one- or two-line problem prompt and ask: "Which pattern, and your one-sentence approach?" (e.g. "Longest contiguous subarray with at most K distinct chars — pattern + one-sentence approach?")
- Complexity — show a short ```python snippet (≤ ~8 lines) and ask: "Time and space?" Pull the snippet from the taxonomy templates or a close variant; ground the answer in the Big-O cheatsheet.
- Template recall — ask: "Sketch the
<pattern>skeleton" (e.g. "Sketch the BFS-shortest-path skeleton — queue, visited, when you mark seen"). The user types a rough skeleton; you check it against the taxonomy template.
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.
- 4d ago First seen · 82 lines · 26 tokens per session scan A 2e80b562beac
coding-drill is a command published in the GitHub repository kirilxd/swe-interview-coach (79 stars, last pushed 1mo ago), licensed MIT. It adds 26 tokens to every session and 1,477 once invoked, about $0.0001 per session on Opus 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 2026-08-30.
Other commands, from other repositories
learn
Force claude-smart to extract learnings from this session now.
recall-save
Generate / overwrite .recall/context.md with Recall's local offline summarizer.
show
Show this project's saved Recall context (.recall/context.md).
init
Bootstrap project context (project.intent.md and project.glossary.json) from an existing codebase using deterministic scan + LLM synthesis + interactive editing. Use --harness to scaffold additional repo-level harness docs.
check
Is the context layer still true? 0-token staleness + broken-citation audit (no LLM, CI-friendly). Add --status for the fast hash-only subset.
save
Save the current session digest. Fans out across every writable memory tier per the house-map. Tier 0 always; Tier 1 if reachable; Tier 2 only if registered writable with auth. Each tier independent — failures degrade gracefully.