القائمة
Sync vendored third-party skills against their upstream repositories, re-apply local patches, and flag new upstream skills. Invoke explicitly with /sync-skills.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
This skill should be used when creating a GitHub pull request for the Composio SDK repository. It generates conventional commit-style PR titles and structured descriptions with Linear issue links, concise bullet points, and optional mermaid diagrams. Triggers on "create PR", "open PR", "make a pull request", or when the user is ready to submit changes for review.
Commit completed units of work as you go instead of batching everything into one commit at the end. Use during any coding session that will produce more than one logical change: after each green test run, when a logical unit is complete, or before starting a risky refactor. Also applies when the user says "commit often", "checkpoint", or "commit as you go".
Help users operate the published Composio CLI to find the right tool, connect accounts, inspect schemas, execute tools, subscribe to trigger events with `composio listen`, script workflows with `composio run`, and call authenticated app APIs with `composio proxy`. Use when the user asks how to do something with `composio`, wants to run a known tool slug, needs to discover a slug with `composio search`, fix a missing connection with `composio link`, inspect tool inputs with `--get-schema` or `--dry-run`, troubleshoot top-level CLI flows, or explicitly needs `composio dev` guidance.
Use when the user wants to design, redesign, shape, critique, audit, polish, clarify, distill, harden, optimize, adapt, animate, colorize, extract, or otherwise improve a frontend interface. Covers websites, landing pages, dashboards, product UI, app shells, components, forms, settings, onboarding, and empty states. Handles UX review, visual hierarchy, information architecture, cognitive load, accessibility, performance, responsive behavior, theming, anti-patterns, typography, fonts, spacing, layout, alignment, color, motion, micro-interactions, UX copy, error states, edge cases, i18n, and reusable design systems or tokens. Also use for bland designs that need to become bolder or more delightful, loud designs that should become quieter, live browser iteration on UI elements, or ambitious visual effects that should feel technically extraordinary. Not for backend-only or non-UI tasks.
Phase-driven workflow for building a feature end to end. Four phases run in order: spec gathers intent and constraints, plan turns the spec into commit-sized units, work executes the plan one commit at a time, and safe-change handles scoped edits where a full spec/plan cycle is overkill. Use when starting new work, when an existing plan needs to be executed, or when making a single change that still needs verification.
Invoke jk, @jkomyno's deterministic personal CLI, for mechanical agent-friendly tasks: read the authenticated GitHub user (whoami), send Telegram notifications or wrap a long command and notify on its exit status (notify), pin and check GitHub Actions refs (actions bump/check), and acquire / index / search per-idea reference corpora under ~/.jk/ideas (idea). Use this skill when running any `jk ...` command, deciding which jk command fits a task, or parsing jk output in a script or hook. It documents the output contract (piped stdout is JSON, a TTY is text, --json forces either) and the exit codes (0 ok, 1 findings, 2 usage, 3 operational).
| name | sync-skills |
| description | Sync vendored third-party skills against their upstream repositories, re-apply local patches, and flag new upstream skills. Invoke explicitly with /sync-skills. |
| disable-model-invocation | true |
Third-party skills are vendored into the dotfiles repository under target/home/.agents/skills/<name>/ and deployed by mise to ~/.agents/skills/<name>/. manifest.json in this skill's directory maps each vendored skill to its upstream repository and in-repo path. This skill keeps the vendored copies in sync with upstream while preserving deliberate local modifications as patch files.
Inspired by dmmulroy's sync-pocock-skills, generalized to many upstreams.
Always operate on the dotfiles source checkout, never on the deployed ~/.agents/skills, so accepted changes land in git:
DOTFILES="${DOTFILES:-$HOME/work/me/dotfiles}"
SYNC_ROOT="$DOTFILES/target/home/.agents/skills/sync-skills"
bash "$SYNC_ROOT/scripts/sync.sh" --keep-upstream
Run sync.sh as above and parse the output sections:
(changed, no patch), (upstream changed, has patch), (patch conflict), (new file upstream), (removed upstream).manifest.json.patches/excluded.txt.--keep-upstream) for reading upstream files.For each CHANGED: entry:
(changed, no patch) / (new file upstream) — safe to take upstream. Run:
bash "$SYNC_ROOT/scripts/apply-upstream.sh" <skill_name> <upstream_skill_dir>
(upstream changed, has patch) — apply-upstream.sh copies upstream and re-applies the stored patch automatically. If it prints PATCH_CONFLICT, read both versions, resolve in the vendored copy, then regenerate the patch with make-patch.sh.
(patch conflict) — same manual resolution: read upstream and ours, merge, regenerate the patch.
(removed upstream) — flag to the user; suggest deleting the file or the whole skill (and its manifest entry).
<upstream_skill_dir> is <clone dir from UPSTREAM_DIRS>/<path from manifest.json>.
When we deliberately diverge from upstream (ours-only edits to a vendored file), store the divergence as a patch so the next sync re-applies it:
bash "$SYNC_ROOT/scripts/make-patch.sh" <skill_name> <rel_path> <upstream_file> <our_file>
Patches live in patches/<skill_name>__<rel_path>.patch with stable upstream/.../ours/... labels, so they do not churn when temp clone paths change.
Frontmatter values that should be preserved without turning into one-line text patches live in patches/local-overrides.json. apply-upstream.sh applies them after copying upstream files and applying patches.
manifest.json with repo and path (the directory containing SKILL.md).apply-upstream.sh <name> <upstream_skill_dir> to vendor it under target/home/.agents/skills/<name>/.~/.claude/skills at the same canonical ~/.agents/skills directory.sync.sh and confirm UPSTREAM_CHANGES is clean (or only intentionally-divergent entries remain).UNPATCHED_PATTERNS is empty or only contains intentionally accepted wording.git -C "$DOTFILES" diff --stat and read the per-skill diffs before reporting; separate metadata-only changes from instruction-body changes.~/.agents/skills update on the next mise dotfiles apply.Report: skills updated, patches created or re-applied, conflicts needing attention, new skills offered/excluded, and the net diff.
sync-skills/
├── SKILL.md # This file
├── manifest.json # skill name -> { repo, path }
├── scripts/
│ ├── sync.sh # Analyse all upstreams vs our vendored copies
│ ├── apply-upstream.sh # Replace vendored copy with upstream + re-apply patches
│ ├── apply-frontmatter-overrides.py # Preserve configured local metadata
│ └── make-patch.sh # Record a local modification as a stable patch
└── patches/
├── excluded.txt # Upstream skills we never offer to vendor
├── local-overrides.json # Frontmatter overrides that are not text patches
└── <skill>__<file>.patch # Local modifications, re-applied on every sync