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 skills/melodic-software/claude-code-plugins/scrutinize-dont-coastnpx skills add melodic-software/claude-code-plugins --skill scrutinize-dont-coastgit 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/scrutinize-dont-coast)<a href="https://agentmods.dev/skills/melodic-software/claude-code-plugins/scrutinize-dont-coast"><img src="https://agentmods.dev/badge/skills/melodic-software/claude-code-plugins/scrutinize-dont-coast.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 | $0.00120 | $0.02042 |
| Opus 5 | $0.00060 | $0.01021 |
| Sonnet 5 | $0.00024 | $0.00408 |
| Haiku 4.5 | $0.00012 | $0.00204 |
Grade A, and why
scrutinize-dont-coast 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 yesterday.
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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Scrutinize, don't coast
A drift corrector for adversarial self-scrutiny. The shared method. Re-anchor,
audit the work in flight, correct forward, report, and the tone that firing this
is not an accusation, lives in
${CLAUDE_PLUGIN_ROOT}/context/re-anchor-audit-correct.md.
Read it; this file adds only what is specific to adversarial self-scrutiny. And
two deliberate deltas this corrector makes to the shared loop.
The discipline this re-anchors
Every other corrector in this plugin re-anchors a content discipline. Research, incumbency, terseness. This one re-anchors a meta discipline: adversarial self-scrutiny of your own recent output. Confidence that the output is sound is not evidence that it is; fluent, momentum-carried work reads as finished long before it has been tested. The lens is not "does this conform to discipline X" but "is this actually right, and would it survive a hostile second look."
Resolve the source of truth per the method doc's ladder: if the consuming project
states a self-scrutiny / verify-before-acting discipline in its own CLAUDE.md or
.claude/rules/, re-anchor THAT. Otherwise re-anchor this portable baseline:
- Confidence is not correctness. Output that flows smoothly and reads as finished has not thereby been tested. Treat "this looks done" as the cue to scrutinize, not the verdict.
- Attack the work, don't defend it. The pass looks actively for what is wrong: a wrong assumption, an unstated leap, a case not handled, a claim not backed, a requirement quietly dropped. Rather than re-reading for reassurance.
- The producing context is the weakest critic. The reasoning that made a mistake plausible is still active, so a hostile look must come from outside it (see the fresh-context rule below).
- A clean pass is a real outcome. Scrutiny that finds the work sound reports clean. The duty is to look hard, not to manufacture a fault.
Three deltas to the shared loop
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.
- yesterday First seen · 144 lines · 120 tokens per session scan A 5a8ed1b57852
scrutinize-dont-coast is a skill published in the GitHub repository melodic-software/claude-code-plugins (15 stars, last pushed today), licensed MIT. It adds 120 tokens to every session and 2,042 once invoked, about $0.0006 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 2026-09-03.
Other skills, from other repositories
parallel-orchestrator
Manage parallel Claude Code workstreams using git worktrees. Use when: splitting large tasks across multiple workers, coordinating parallel development, monitoring worker progress, integrating completed work, analyzing work item documents (code reviews, issue lists). Triggers: parallel, orchestrator, worktrees…
parallel-worker
Execute focused implementation tasks in a parallel workflow. Use when: working on assigned files in a worktree, making checkpoint commits, signaling dependencies or blockers, completing orchestrator-assigned tasks. Triggers: worker, checkpoint, worktree, assigned scope, commit prefix, parallel task.
generate-tree-structure
For tree/maze generation: spanning trees, random mazes, graph coverage. Uses frontier-based exploration with configurable traversal order.
optimize-local-search
For NP-hard optimization: TSP, scheduling, assignment problems. Uses greedy construction + local improvement (2-opt, hill climbing).
solve-constraint-puzzle
For constraint satisfaction: Sudoku, scheduling, N-queens, logic puzzles, SAT-like problems, assignment problems. Uses propagate-then-search pattern.
use-class-for-state
For complex state: encapsulate multiple interacting variables, stateful algorithms, backtracking search state.