add-skill
Install a pinned skill extension on demand (/add-skill <id>), or list core packs and extensions
Apply the fixes the user approved on a review sheet, re-verify each one, and update the sheet
> /plugin marketplace add solanabr/ai-kit > /plugin install solana-ai-kit@stbr
How it fires
How this command gets triggered: by you, by Claude, or both.
/fix-waveContext preview
What this command does when you run it.
Apply the fixes the user approved on a review sheet, re-verify each one, and update the sheet
description: "Apply the fixes the user approved on a review sheet, re-verify each one, and update the sheet"
Apply the fixes the user approved on a `/review-round` sheet: the one named in `$ARGUMENTS`, else the newest `.claude/context/review-*.md`. Only what was approved, and never a push or a merge.
An `ask` nobody answered, a `defer`, and any item on a `--candidates` sheet they did not pick stay as they are. A timeout, an away notice, or a reply asking to wait or explain leaves an item unanswered.
1. **Record the answers** that settle a question (an `ask` they answered, a default they overruled): `bash .claude/bin/decisions.sh add --by you --kind decision --question '<the item>' --decision '<what they chose>' --words '<their words, verbatim>'`, adding `--date YYYY-MM-DD` when they said it on another day. Single-quote every value, writing an apostrophe as `'\''`. If the helper is missing or fails, say so and keep their words quoted in the sheet. 2. **Check the head.** The sheet names the commit it reviewed. If the branch has moved since (`git rev-parse <branch>`), re-check each approved item against the new code before touching it. 3. **Fix in parallel, by owned files,** on the branch whose head was reviewed, in its own worktree if it is not checked out. When that head is the default branch itself (a `since <date>` audit, or a range ending there), fix on a new branch from it, named by the kit's branch convention. Group the approved items by the files they touch: groups that share no file run at once, fanned out the way [/implement](implement.md) does (a worktree each from the sheet's head, one integrator, WIP commits, no merges); a single group is fixed in this session. 4. **Re-check each fix** with the evidence that proved the finding: the reproduction must stop reproducing, and the checks that cover the touched files must pass, where zero tests or SKIP counts as a failure. A finding without a reproduction gets a test that would have caught it, or a re-check by a fresh agent with its lens. 5. **Stop on an item** when its re-check fails after two different fixes. It goes back on the sheet as `ask`, with both attempts and their results. 6. **Commit** per item or per group, the message naming the sheet items it closes. A problem found while fixing becomes a new numbered item on the sheet, not an unasked extra change. 7. **Update the sheet:** each item `fixed` (commit, command and result), `failed` (why) or `skipped` (why).
Per item: its number, status, commit, and the re-check command with its result. Then the head before and after the wave: the fixes moved it, so the sheet no longer covers the head. Run `/review-round <old head>..<new head>` before `/ship`, which accepts only a review of the commit it merges.
Claude Code, set up to ship on Solana. Not programs alone — the whole lifecycle: validating the idea, building the program and the app around it, shipping it, and the launch and growth work that follows.
Install a pinned skill extension on demand (/add-skill <id>), or list core packs and extensions
Short for /adversarial: run every task through proposer and adversary pairs, no triage
Put a task's load-bearing claims and key decisions through a proposer and an adversary, with proof
Audit infra security: secrets, deps, CI/CD, webhooks, AI/skill files
Audit Solana program code for exploitable bugs and write a findings report
Compare per-instruction CU with the stored baseline to catch regressions