managing-parsers

SkillDev tools

Use when the user needs to manage the tree-sitter parser cache — prefetch parsers for offline or CI runs, list what is downloaded, inspect a language, find the cache directory, or clean it. Covers `ts-pack download`, `list`, `info`, `cache-dir`, `clean`, and `init`.

Available today. Use it from your connected AI after setup.

Connect ahel once, and every AI you use reads what you have installed.

Then ask your AI: use the managing-parsers skill

What this skill tells your AI

The instructions your AI receives, as published by xberg-io/tree-sitter-language-pack in plugin/skills/managing-parsers/SKILL.md and read by ahel’s review.

Managing parsers

Parser libraries download on demand the first time a language is used and are cached on disk. Manage that cache explicitly for offline work, CI prefetch, reproducible builds, or to reclaim space.

Note: parser cache management (download, clean, init) is done via the ts-pack CLI directly. The OpenCode tools expose only parse/process/info.

Prefetch (offline / CI)

ts-pack download python rust typescript   # specific languages
ts-pack download --all                    # every available language
ts-pack download --groups all             # by group
ts-pack download python --fresh           # clean cache first, then download

The manifest currently defines exactly one group, all. Group names are manifest data, not a fixed list — enumerate them with manifest_groups() rather than hardcoding. Run download in your CI setup step so later parse/process calls hit the cache instead of the network.

Inspect

ts-pack list                 # all available languages
ts-pack list --downloaded    # only languages currently cached
ts-pack list --manifest      # everything in the remote manifest
ts-pack list --filter script # filter by name substring

ts-pack info python          # known? downloaded? cache path?
ts-pack cache-dir            # print the effective cache directory

Clean

ts-pack clean           # prompts for confirmation
ts-pack clean --force   # skip the prompt (use in scripts/CI)

clean removes all cached parser libraries; they re-download on next use.

Config file (init)

ts-pack init writes a language-pack.toml to the current directory pinning a cache location and/or a language set. This writes a file to the working directory.

ts-pack init --languages python,rust --cache-dir .ts-pack-cache

Resulting language-pack.toml:

cache_dir = ".ts-pack-cache"
languages = ["python", "rust"]

With no flags, init writes a commented template. When --languages is given, init also triggers the download for those languages. (init takes only --languages and --cache-dir — for group-based prefetch use ts-pack download --groups <names>.) Commit language-pack.toml to pin the parser set across a team or CI.

Typical CI flow

ts-pack init --languages python,rust,typescript --cache-dir .ts-pack-cache
# ... later steps run parse/process offline against the warm cache ...
ts-pack list --downloaded   # assert the expected parsers are present

When to reach for the other skills

Once parsers are cached, use parsing-source to get syntax trees and extracting-code-structure to pull metadata. info/list here also serve the language-support checks described in detecting-languages.

Signals

GitHub stars
465
Forks
68
Last commit
Sep 2026
Advanced
Catalog kind
skill
Gateway key
managing-parsers
Source
github.com/xberg-io/tree-sitter-language-pack