* feat(#778): cross-runtime command enrichment (Gemini {{args}}/!{}, Qwen priority) Enrich the installer's per-runtime command/skill generators with native, verified, additive fields: - Gemini CLI: map Claude's $ARGUMENTS -> Gemini's {{args}} in generated TOML commands so typed arguments interpolate; inject live .planning/STATE.md into /gsd:progress via a fixed, injection-safe !{cat .planning/STATE.md 2>/dev/null} shell block (no interpolated input). - Qwen Code: emit the optional numeric `priority` field on main-loop skills so the most-used workflows sort first in the /skills list (higher = earlier per the Qwen skills spec; the issue's inverse numbering was corrected). OpenCode per-command model/agent/subtask/variant enrichment was evaluated and intentionally not implemented: `model` reintroduces the #1156 ProviderModelNotFoundError regression for non-Anthropic providers (the converter deliberately strips model:), `subtask`/`agent` change execution semantics for GSD's interactive commands, and `variant` is not in the OpenCode command schema. Schemas verified against primary docs (Gemini custom-commands, Qwen skills, OpenCode commands/skills). Adds tests/enh-778-* and how-to + USER-GUIDE docs. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> * chore(#778): set changeset PR number to 825 Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Tom Boucher
@@ -2108,6 +2108,40 @@ function skillFrontmatterName(skillDirName) {
|
||||
return skillDirName;
|
||||
}
|
||||
|
||||
/**
|
||||
* Qwen Code skills accept an optional numeric `priority` frontmatter field.
|
||||
* Per the Qwen skills spec (qwen-code/docs/users/features/skills.md, verified
|
||||
* #778): HIGHER values sort EARLIER in the `/skills` TUI listing (omitted ≈ 0;
|
||||
* negatives sort below unset). It affects ONLY the `/skills` list order —
|
||||
* slash-command completion and the `/help` view stay alphabetical.
|
||||
*
|
||||
* We assign descending priorities to GSD's main-loop commands so the most-used
|
||||
* workflow skills surface first; utility skills are deliberately left unset
|
||||
* (default 0) and sort below.
|
||||
*
|
||||
* NOTE: the #778 issue body proposed the INVERSE numbering (plan-phase: 10,
|
||||
* utilities: 90+). The verified spec shows that would BURY the core loop below
|
||||
* utilities, so we implement the spec-correct direction (core = high) instead.
|
||||
* Keyed by command stem (skill dir is `gsd-<stem>`).
|
||||
*/
|
||||
const QWEN_SKILL_PRIORITY = Object.freeze({
|
||||
'new-project': 100,
|
||||
'discuss-phase': 95,
|
||||
'plan-phase': 90,
|
||||
'execute-phase': 85,
|
||||
progress: 80,
|
||||
'verify-work': 75,
|
||||
phase: 70,
|
||||
review: 65,
|
||||
ship: 60,
|
||||
config: 55,
|
||||
surface: 50,
|
||||
'resume-work': 45,
|
||||
'pause-work': 40,
|
||||
help: 35,
|
||||
update: 30,
|
||||
});
|
||||
|
||||
/**
|
||||
* Convert a Claude command (.md) to a Claude skill (SKILL.md).
|
||||
* Claude Code is the native format, so minimal conversion needed —
|
||||
@@ -2151,6 +2185,18 @@ function convertClaudeCommandToClaudeSkill(content, skillName, runtime = null, c
|
||||
// Track GSD's package version so Hermes' skill_view() reports a stable
|
||||
// identifier per install.
|
||||
if (runtime === 'hermes') fm += `version: ${yamlQuote(pkg.version)}\n`;
|
||||
// #778 (b) — Qwen-only numeric priority for /skills ordering. Scoped to qwen
|
||||
// so Claude/Hermes skill frontmatter is unchanged (they ignore the field, but
|
||||
// we keep their output byte-stable). skillName is the `gsd-<stem>` dir name.
|
||||
if (runtime === 'qwen') {
|
||||
const stem = typeof skillName === 'string' && skillName.startsWith('gsd-')
|
||||
? skillName.slice(4)
|
||||
: skillName;
|
||||
const priority = Object.prototype.hasOwnProperty.call(QWEN_SKILL_PRIORITY, stem)
|
||||
? QWEN_SKILL_PRIORITY[stem]
|
||||
: undefined;
|
||||
if (typeof priority === 'number') fm += `priority: ${priority}\n`;
|
||||
}
|
||||
if (argumentHint) fm += `argument-hint: ${yamlQuote(argumentHint)}\n`;
|
||||
if (agent) fm += `agent: ${agent}\n`;
|
||||
// #769: emit context: and effort: when present so the runtime can honour
|
||||
@@ -5875,7 +5921,7 @@ function convertSlashCommandsToGeminiMentions(content) {
|
||||
});
|
||||
}
|
||||
|
||||
function convertClaudeToGeminiMarkdown(content, { isCommand = false } = {}) {
|
||||
function convertClaudeToGeminiMarkdown(content, { isCommand = false, commandName = null } = {}) {
|
||||
// Apply Gemini-specific slash command namespacing
|
||||
let converted = convertSlashCommandsToGeminiMentions(content);
|
||||
// Gemini CLI does not expose Claude's AskUserQuestion tool. Convert body
|
||||
@@ -5887,8 +5933,9 @@ function convertClaudeToGeminiMarkdown(content, { isCommand = false } = {}) {
|
||||
converted = stripSubTags(converted);
|
||||
|
||||
if (isCommand) {
|
||||
// Convert to Gemini TOML format
|
||||
converted = convertClaudeToGeminiToml(converted);
|
||||
// Convert to Gemini TOML format (threads the command name so per-command
|
||||
// enrichment — e.g. the #778 live-state injection — can target a command).
|
||||
converted = convertClaudeToGeminiToml(converted, { commandName });
|
||||
}
|
||||
|
||||
return converted;
|
||||
@@ -6392,7 +6439,16 @@ function convertClaudeCommandToKiloSkill(content, skillName) {
|
||||
* @param {string} content - Markdown file content with YAML frontmatter
|
||||
* @returns {string} - TOML content
|
||||
*/
|
||||
function convertClaudeToGeminiToml(content) {
|
||||
function convertClaudeToGeminiToml(content, { commandName = null } = {}) {
|
||||
// #778 (c) — Gemini {{args}} interpolation. Claude's $ARGUMENTS placeholder
|
||||
// maps to Gemini's {{args}} so inline argument references interpolate into the
|
||||
// command body instead of being emitted as a dead literal. Applied before
|
||||
// frontmatter parsing so every return path benefits (a command's frontmatter
|
||||
// never contains $ARGUMENTS, so this is body-only in practice). Gemini injects
|
||||
// {{args}} as typed outside shell blocks; we never place it inside a !{...}
|
||||
// block, so there is no shell-escaping/injection interaction.
|
||||
content = content.replace(/\$ARGUMENTS\b/g, '{{args}}');
|
||||
|
||||
// Check if content has frontmatter
|
||||
if (!content.startsWith('---')) {
|
||||
return `prompt = ${JSON.stringify(content)}\n`;
|
||||
@@ -6404,7 +6460,27 @@ function convertClaudeToGeminiToml(content) {
|
||||
}
|
||||
|
||||
const frontmatter = content.substring(3, endIndex).trim();
|
||||
const body = content.substring(endIndex + 3).trim();
|
||||
let body = content.substring(endIndex + 3).trim();
|
||||
|
||||
// #778 (c) — Gemini !{...} dynamic-output injection for the situational
|
||||
// `progress` command (GSD's status/dashboard surface). Inject the live
|
||||
// .planning/STATE.md so the model sees current project state without relying
|
||||
// on session memory.
|
||||
//
|
||||
// SECURITY: the shell command is a FIXED `cat` with NO interpolated user
|
||||
// input — no {{args}} appears inside the block — so there is no
|
||||
// shell-injection vector. Gemini still shows its standard per-invocation
|
||||
// confirmation dialog (verified behavior). `2>/dev/null` keeps an
|
||||
// uninitialized project (missing STATE.md) from injecting stderr noise.
|
||||
// Braces inside the block are balanced (none present), per Gemini's parser
|
||||
// requirement. The append happens AFTER the {{args}} mapping above so the
|
||||
// injected block can never accidentally carry interpolated arguments.
|
||||
if (commandName === 'progress') {
|
||||
body += '\n\n## Live project state\n'
|
||||
+ 'Current contents of `.planning/STATE.md` '
|
||||
+ '(empty if the project is not yet initialized):\n\n'
|
||||
+ '!{cat .planning/STATE.md 2>/dev/null}\n';
|
||||
}
|
||||
|
||||
// Extract description from frontmatter
|
||||
let description = '';
|
||||
@@ -7431,8 +7507,11 @@ function copyWithPathReplacement(srcDir, destDir, pathPrefix, runtime, isCommand
|
||||
: convertClaudeToOpencodeFrontmatter(content);
|
||||
fs.writeFileSync(destPath, content);
|
||||
} else if (isGemini) {
|
||||
// Apply Gemini-specific Markdown transformations (slash commands, TOML)
|
||||
const processed = convertClaudeToGeminiMarkdown(content, { isCommand });
|
||||
// Apply Gemini-specific Markdown transformations (slash commands, TOML).
|
||||
// #778: thread the command name (file stem) so per-command TOML
|
||||
// enrichment (live-state injection) can target a specific command.
|
||||
const geminiCommandName = isCommand ? entry.name.replace(/\.md$/, '') : null;
|
||||
const processed = convertClaudeToGeminiMarkdown(content, { isCommand, commandName: geminiCommandName });
|
||||
const finalPath = isCommand ? destPath.replace(/\.md$/, '.toml') : destPath;
|
||||
fs.writeFileSync(finalPath, processed);
|
||||
} else if (isCodex) {
|
||||
|
||||
Reference in New Issue
Block a user