From rocket
Produce a precise change plan (file:line + short prose + why) for a human developer to apply themselves, instead of editing files. Manual-only — invoked by the user via "/rocket:myself" or "/myself".
How this skill is triggered — by the user, by Claude, or both
Slash command
/rocket:myselfThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
The user wants to apply the changes themselves. Your role is to act as a high-precision pointer: identify exactly where each change must happen and explain — concisely — what to do and why. The human will write the code.
The user wants to apply the changes themselves. Your role is to act as a high-precision pointer: identify exactly where each change must happen and explain — concisely — what to do and why. The human will write the code.
This is a deliberate two-actor split:
A single structured plan with this shape:
One sentence restating what the user wants to achieve. Skip if obvious.
Up to 3 bullets surfacing non-obvious facts the human needs before starting (e.g. an existing helper to reuse, a constraint, a related file that must stay in sync). Omit entirely if there is nothing useful to add.
A numbered list. Each entry:
N. `path/to/file.ext:LINE` — <concise action in prose>
Why: <one short line>
LINE is the exact line number (or range 42-58) where the change happens. Verify it by reading the file; do not guess.Up to 3 bullets for side effects, follow-ups, or things the human should verify after applying the plan (tests to run, types to regenerate, callers to update if you are not listing them as changes, config to reload). Omit if there is nothing to flag.
If you cannot pinpoint a line because the change is genuinely ambiguous (multiple plausible call sites, the user's intent is unclear, the relevant code does not exist yet), say so explicitly in the entry — e.g. path/to/file.ext:? with a short note on what to confirm — rather than inventing a number. Ask a focused question at the end of the plan only for blockers; do not chain questions.
User: "add a retry with exponential backoff around the Stripe charge call"
### Changes
1. `src/payments/stripe-client.ts:88` — wrap the `charge` call in a retry helper that retries on network/5xx errors, max 3 attempts, base delay 200ms doubling each time. Pull the retry helper into a separate function in the same file rather than inlining.
Why: keeps the retry policy visible and unit-testable next to the only caller.
2. `src/payments/stripe-client.ts:12` — add `MAX_RETRIES` and `BASE_DELAY_MS` as module constants.
Why: avoids magic numbers and lets tests override via re-export.
3. `src/payments/__tests__/stripe-client.test.ts:?` — add a test that simulates two transient failures followed by success and asserts a single resolved charge.
Why: regression guard for the retry path.
### Out of scope / watch out
- The webhook handler in `src/payments/webhooks.ts` already retries idempotency-key-keyed requests; do not double-retry there.
- Verify Stripe SDK errors expose `.statusCode` in the version pinned in `package.json` before relying on it.
npx claudepluginhub digivorefr/roc --plugin rocketGuides creation, editing, and verification of skills for AI coding agents using test-driven development with subagent scenarios. Use when authoring or debugging skills.