sattva2020
- 28 skills
- 0 followers
- 2 days ago last updated
- ▌ Transfer · sattva2020Keystone: reuse the lessons of another of your projects. Reads (never writes) its .claude/rules, evolve log, fix-plan Prevention sections and stack line, keeps only what fits this project's stack, marks every proposal with its source, and writes the accepted ones into .claude/rules/ after you see the diff and say yes. Use when starting a project similar to an older one, or to avoid mistakes already paid for elsewhere.
- ▌ Dockerize · sattva2020 bundleKeystone: analyze the project and generate Docker configuration — multi-stage Dockerfile (dev/prod), compose.yml, compose.override.yml (dev), hardened compose.production.yml, .dockerignore, deploy scripts — or enhance / audit an existing setup, with a production security score. Reverse proxy is an option, not a default. Nothing is written until you have seen the files and said yes. Use for 'dockerize', 'add docker', 'docker compose', 'containerize', 'setup docker'.
- ▌ Implement · sattva2020 bundleKeystone: execute an approved plan task by task — requirement consistency gate, checkbox after every task, resume after /clear, plan drift edited in the same commit, literal BLOCKED:, line-by-line evidence report. Use only via explicit /keystone:implement [plan path].
- ▌ Reference · sattva2020 bundleKeystone: turn URLs, documents or files into a project reference in docs/references/[slug].md. Library and SDK docs come from Context7 first; the reference keeps only project-specific conclusions, prices and limits, and live checks with date and source URL — no copies of docs Context7 already serves. Use for 'make a reference', 'document this API for the agent', 'save these pricing limits'.
- ▌ Architecture · sattva2020 bundleKeystone: recommend an architecture pattern from the project's real stack and code, check the existing structure against it, and write docs/architecture.md (folder structure, dependency rules in a form /keystone:verify can grep, anti-patterns) after you confirm the diff. Re-running updates the file, never overwrites silently. Use for 'which architecture', 'document the architecture', after /keystone:setup, or before a structural refactor.
- ▌ Setup · sattva2020 bundleKeystone: set up agent context for a project — analyze the stack, write a short CLAUDE.md, .claude/rules/base.md, a project .mcp.json (secrets only as ${VAR}) and recommend Claude Code plugins. Existing project, new project from a description, or an empty folder. Every write is shown as a diff and needs an explicit yes; a re-run adds only what is missing and never overwrites. Use for 'set up the project', 'configure AI context', 'init', 'what MCP / plugins do I need'.
- ▌ Start · sattva2020Keystone: size a change as S, M or L and route it — S (diff in one sentence) → code + check, no intent, no plan; M (one clear approach) → short plan with 3-line Risks; L (several decisions, data, security, other teams' modules) → intent → plan → implement → verify. Picks the model per stage, caps exploration, and asks for the cost line in the report. Use at the start of any task: 'start', 'what level is this', 'how should we do this'.
- ▌ Evolve · sattva2020Keystone: turn the lessons of finished tasks into project rules. Reads fix plans (Root cause, Prevention), verify FAIL and Found lines, Implementation notes deviations, saved review and security findings and docs/notes; finds recurring patterns and stack pitfalls; proposes rules for .claude/rules/ — written only after you see the diff and say yes. Processed sources are logged in docs/keystone/evolve-log.md so a re-run skips them. `add [rule text]` writes one rule you dictate (dedupe, topic file, paths:) after the same diff and yes. Use for 'evolve', 'learn from our fixes', 'what keeps going wrong', after several fixes, and for 'add rule', 'remember this convention', 'always do X'.
- ▌ Intent · sattva2020Capture a change as intent.md — what and why, in the originator's words, decisions confirmed via /keystone:grill, every line with an ID (I-1…) — and have the human sign it off. The first file of the Keystone chain intent → plan → code → verify (Anthropic's AI-native SDLC playbook, adapted: the plan is the spec). Use for 'intent', 'capture the idea', 'what exactly do we want', or before planning an L-level change.
- ▌ Review · sattva2020 bundleKeystone: read-only code review of staged changes, a pull request (gh pr diff) or everything since a git ref — correctness, security, performance, project rules and architecture. Every finding carries severity, file:line, a Repro and a fix; an independent validator in a clean context ALWAYS re-checks them (keep / modify / drop, severity changes), unproven ones go to Unconfirmed. `rules [ref]` checks only the project's MUST / MUST NOT rules over the diff, without a plan (S changes), one line per rule with every violating file:line, plus suggested rule updates for /keystone:evolve. Use for 'review code', 'check my code', 'review PR', 'is this code okay', 'check the rules', and on L changes after /keystone:verify.
- ▌ Verify · sattva2020 bundleKeystone: verify finished work against its plan and intent — runs build/test/lint, then an independent verifier agent in a clean context gives every plan task, Reconciliation row and intent line ok / missing / wrong / not proven / BLOCKED, flags unrequested work, and returns PASS, PASS-UNPROVEN or FAIL. A checked box is not evidence. Writes verify.md; on PASS sets Status verified. Use after implementation on M and L changes, or for 'verify', 'did we miss anything', 'is it really done'. `/keystone:verify accept` records a human decision on unproven rows.
- ▌ Pentest · sattva2020 bundleKeystone: confirm the findings of /keystone:security against your own running app — a scope gate first (your own localhost / staging target only, production observation-only), passive checks (headers, CORS, cookies, TLS, reachable endpoints), then one safe request per finding: Confirmed with a PoC, Not reproduced, or Needs manual. Report in Sentry's form with Refs and severity. Never brute force, DoS, bypass protections or touch third-party services. Use for 'pentest', 'confirm the security findings', 'is this really exploitable'.
- ▌ Roadmap · sattva2020Keystone: create or update docs/roadmap.md — a strategic checklist of milestones with stable IDs (M-1, M-2, …) that intents can reference (Roadmap: M-[n]). Modes: create, update, check progress (reads task folders to propose ticks). Writes only after you see the diff and say yes. Use for 'roadmap', 'milestones', 'what to build next'.
- ▌ Optimize · sattva2020 bundleKeystone: make code faster or simpler on evidence — measure before and after with the same command, keep a change only if the numbers improve beyond noise (perf), or simplify without changing behaviour with tests green before and after (simplify). Reports before → after with a median and a spread, reverts a change that is within the noise, never touches production. Use for 'optimize', 'make it faster', 'this is slow', 'simplify this code', or a performance finding from /keystone:review.
- ▌ QA Check · sattva2020 bundleKeystone: executes the test cases written by /keystone:qa — human-guided (asks the user case by case) or agent mode (browser via the built-in browser tools or Playwright MCP, CLI, tests, API, file checks) — and records pass / fail / blocked per case with evidence in [task folder]/qa/qa-check.md. A case without evidence is not pass. Reusable browser replay = Playwright test scripts in [task folder]/qa/replay/. Use for 'run the QA', 'walk through the test cases', 'verify the cases', 'rerun QA'.
- ▌ Security · sattva2020 bundleKeystone: security audit on OWASP Top 10:2025, LLM Top 10 2026 / prompt injection, agent and MCP configuration, supply chain, IaC (Dockerfile, Kubernetes, Terraform) and race conditions — whole project, a path, a category or the diff of a task. Every finding has severity, file:line, how to exploit it and a fix; ignored items live in docs/keystone/security-ignores.md with reason, date and author; ends with a machine-readable line `Security gate: pass | warn | fail` (fail only on an unignored critical/high finding). Read-only for code. Use for 'security check', 'is this secure', 'before deploy', auth/payments/user-input changes, or `ignore [id] [reason]`.
- ▌ Sessions · sattva2020 bundleKeystone: sort out the child sessions a main session started — archive the finished ones, send each unfinished idle one the single step an agent can do itself, and list the rest as 'waiting for you' with a one-line reason. The decision is mechanical (git + plan files, by scripts/triage.mjs); only the family of this session (linked: true) is touched. Runs by itself when the app reports on side sessions (a hook adds the request), or on demand. Use for 'archive finished sessions', 'what are my sub-sessions doing', 'clean up side sessions'.
- ▌ CI · sattva2020 bundleKeystone: analyze the project and generate a CI/CD pipeline (GitHub Actions or GitLab CI) — lint, static analysis, tests with service containers, build, security — or enhance / audit an existing one against best practices. Commands come from the tools the project actually has. Nothing is written until you have seen the files and said yes. Use for 'ci', 'setup ci', 'github actions', 'gitlab ci', 'pipeline'.
- ▌ QA · sattva2020 bundleKeystone: QA workflow for testing a feature or task implementation — change summary, test plan and concrete manual test cases, written to [task folder]/qa/. Input is a Keystone plan path, a git ref, or nothing (staged/working diff). With a plan, cases cite intent IDs (I-n) and the plan's `Done means:`. Use for 'test this', 'write test plan', 'what should I test', 'QA this change'; run the cases with /keystone:qa-check.
- ▌ Ask · sattva2020Keystone: answer a question only on evidence — 100/100 confidence from files, command output, docs or verified sources, otherwise INSUFFICIENT INFORMATION with exactly what is missing. No guessing, no code changes. Use for 'be 100% sure', 'no hallucinations', 'only if verified', versions, prices, 'does X support Y now', security-adjacent questions.
- ▌ Fix · sattva2020Keystone: fix a bug regression-first — reproduce, failing test, smallest fix, same check green; root cause and prevention recorded in the plan. Fix now (S) or plan first (M). Use only via explicit /keystone:fix [bug description or error], or /keystone:fix verify to fix every finding of the last FAIL verify.md in one pass.
- ▌ Docs · sattva2020 bundleKeystone: generate and maintain project documentation — a lean README landing page plus topic pages in docs/, with navigation, consolidation of scattered *.md files (only with your yes, nothing deleted automatically) and a review against a checklist. `update [files|topics]` refreshes only the pages a change touched (called by /keystone:implement when Settings say docs — yes). Use for 'create docs', 'write documentation', 'update docs', 'generate readme', 'document project'.
- ▌ Plan · sattva2020 bundleKeystone: plan an M or L change into docs/keystone/[date]-[slug]/plan.md — reconnaissance, Requirements Reconciliation Gate against intent.md (I-n) and project rules, Rules check, Risks, tasks with files and checks, a pre-approval interrogation round, and on L an independent /keystone:check. Settings are inferred from the repo, never asked. Explicit /keystone:plan only.
- ▌ Check · sattva2020 bundleKeystone: independent check of a plan before Approve — deep codebase analysis, a fresh validator, and a line for every intent line (I-1…) with Coverage % and PASS/FAIL. Finds missing or vague tasks, wrong dependencies, duplicates, unrequested work and project-rule violations; edits the plan only after the human confirms. Never edits code. Use before Approve on an L change, after a retrospective intent, or when intent.md changed after the plan.
- ▌ Grill · sattva2020Keystone: interview the user relentlessly, round by round, over the whole decision tree — implementation, UX, edge cases, failures, data, security, trade-offs, what done means — until every branch is visited and nothing is silently assumed. Facts are found by the agent, decisions are made by the user. Writes no file; returns a confirmed, numbered decision list (D1…) to the caller (/keystone:intent, /keystone:plan, /keystone:fix) or to the user. Use for 'grill me', 'interview me', 'stress-test this idea', or when open decisions depend on each other.
- ▌ Harness Audit · sattva2020 bundleAudits a codebase for "AI-agent harness" maturity across six layers (context engineering, tools & MCP, memory & state, tests & evals, observability, guardrails & approvals). Scores each layer, flags AGENTS.md/CLAUDE.md drift, gives prioritized fixes. Use for "audit harness", "оцени харнес", "check agent readiness", "is this project agent-ready", "проверь готовность к AI-агентам".
- ▌ Dokploy API MCP · sattva2020 bundleDeploy and manage applications on Dokploy (self-hosted PaaS). Use when deploying apps, managing Docker Compose services, databases (PostgreSQL, MySQL, MariaDB, MongoDB, Redis), configuring domains, SSL, backups, notifications, rollbacks, scheduled tasks, environments, organizations, SSO, preview deployments, patches, Docker Swarm clusters, running migrations, or troubleshooting Dokploy deployments. Covers tRPC API (~540 endpoints), CLI, MCP server (~540 tools), and common pitfalls with Next.js, Docker, and Traefik.
- ▌ Auth Implementation · sattva2020 bundleImplement authentication and authorization in web applications. Use when adding login, signup, sessions, JWT tokens, OAuth, SSO, API key auth, role-based access control (RBAC), permissions, protected routes, middleware guards, password hashing, MFA/2FA, refresh token rotation, CSRF protection, or integrating auth providers like NextAuth/Auth.js, Better Auth, Lucia, Clerk, Auth0, Supabase Auth, Firebase Auth, Passport.js. Covers Next.js, Express, Fastify, Django, FastAPI, Laravel, Go. Handles database schema design, security hardening, and common auth pitfalls (token storage, session fixation, CORS).