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 waazy-w/claude-phantom --skill crash-recoverygit clone --depth 1 https://github.com/waazy-w/claude-phantomWrote 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/waazy-w/claude-phantom/crash-recovery)<a href="https://agentmods.dev/skills/waazy-w/claude-phantom/crash-recovery"><img src="https://agentmods.dev/badge/skills/waazy-w/claude-phantom/crash-recovery/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/waazy-w/claude-phantom/crash-recovery"><img src="https://agentmods.dev/badge/skills/waazy-w/claude-phantom/crash-recovery.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 | $0.00055 | $0.01788 |
| Opus 5 | $0.00028 | $0.00894 |
| Sonnet 5 | $0.00011 | $0.00358 |
| Haiku 4.5 | $0.00006 | $0.00179 |
Grade C, and why
crash-recovery scanned grade C with 2 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 12d 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.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
2. NEVER run `rm -r`, `rm -rf`, `rmdir`, `sudo`, `chmod -R`, `chown`, `dd`, `mkfs`, `docker`, `kubectl`, or kill processes. Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
4. NEVER use the network (`curl`, `wget`, `nc`, `ssh`, `scp`, `WebFetch`, `WebSearch`). How it starts
The opening of the file, as written. The whole thing — 77 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Crash recovery procedure
You are recovering a crashed command in someone else's repository. Follow every phase in order. Every hard rule applies in every phase. Evidence beats intuition: never guess when you can read.
Hard rules (always)
- NEVER run
git checkout,git switch,git reset,git stash,git rebase,git merge,git commit,git push,git clean,git branch -D. Phantom owns git. Read-only git (git diff,git log,git status,git show) is fine. - NEVER run
rm -r,rm -rf,rmdir,sudo,chmod -R,chown,dd,mkfs,docker,kubectl, or kill processes. - NEVER install, remove, or update packages (
npm install,npm i,npm ci,yarn add,pnpm add,pip install, ...). NEVER edit lockfiles. - NEVER use the network (
curl,wget,nc,ssh,scp,WebFetch,WebSearch). - NEVER run migrations or touch database schemas (
migrate,prisma db,knex,sequelize db,drizzle-kit push,DROP TABLE,TRUNCATE). - NEVER read or edit never-touch paths (the list is injected below the procedure). NEVER edit CI/workflow files (
.github/**,.gitlab-ci.yml,Jenkinsfile,.circleci/**),.gitignore, or anything under.git/. - NEVER disable, delete, or skip existing tests. NEVER loosen an assertion. NEVER add
skip,only,todo,xit,xdescribe, or@pytest.mark.skip. NEVER change a test's expectation to match broken behavior. - NEVER refactor, reformat, rename, reorder imports, bump versions, change dependencies, or "clean up" code you were not asked to fix.
- Stay inside the repository root. Do not write anywhere outside it except the report path you are given.
- When a rule and the goal conflict, the rule wins: stop and write a PARTIAL report.
Phase 0 - Orient
- Read the crash context you were given in full: command, exit summary, error line, stack trace, hint files, recent output, git status, package scripts, test command.
- List candidate files from the trace, innermost project frame first. Ignore
node_modules,node:,internal/frames. - Read every candidate file with the Read tool. Read the callers of the failing function (Grep for its name). Read the nearest existing test file to learn conventions.
- Identify the test runner from
package.jsonscripts and existing test files (node:test, jest, vitest, mocha, pytest, go test, cargo test). Note the exact command that runs ONE test file. - Do not proceed until you can point at
file:linewhere the failure originates.
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.
- 12d ago First seen · 77 lines · 55 tokens per session scan C 4166dbc66af3
crash-recovery is a skill published in the GitHub repository waazy-w/claude-phantom (4 stars, last pushed 20d ago), licensed MIT. It adds 55 tokens to every session and 1,788 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 2 findings (recursive force delete, makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
fix-bug
Diagnoses and fixes bugs using a test-first workflow (reproduce, diagnose, fix). Use this skill whenever the user reports a bug, describes unexpected behavior, or asks to investigate or fix a defect. Even if the user says "something's broken", "this isn't working", "fix this bug", or "why does X happen", load this…
debugging
Use when debugging bugs, test failures, or unexpected behavior. Triggers: 'why isn't this working', 'this doesn't work', 'X is broken', 'something's wrong', 'getting an error', 'exception in', 'stopped working', 'regression', 'crash', 'hang', 'flaky test', 'intermittent failure', or when user pastes a stack…
ralphctl-debugging-and-error-recovery
Systematic root-cause debugging. Use when tests fail, builds break, or behaviour does not match expectations. Follow stop-the-line → reproduce → localize → reduce → root-cause → guard-with-regression-test → verify, not guessing; the reproduction and regression steps follow the same red-green discipline as…
fix-bug
Investigates and fixes a reported bug. Reads relevant code, identifies the root cause, proposes a fix, and adds or updates tests to prevent regression. Invoked when the user describes unexpected behavior, an error, a crash, or a failing test.
regression-test-gen
Suggest regression coverage for changed stack behavior.
Vitest Mock Isolation Review
Detects Vitest test-isolation hazards: unrestored vi.spyOn/vi.mock without afterEach cleanup or restoreMocks config, un-awaited resolves/rejects assertions, and shared mutable module state across tests.