feat: agent skills

[uinstinct]

Description

This PR introduces agent skills to the vscode/jetbrains extension.

References:
About agent skill: https://platform.claude.com/docs/en/agents-and-tools/agent-skills/overview
Integrating agent skill: https://github.com/agentskills/agentskills/blob/main/docs/integrate-skills.mdx

resolves CON-5136

AI Code Review

• Team members only: AI review runs automatically when PR is opened or marked ready for review
• Team members can also trigger a review by commenting @continue-review

Checklist

• [] I've read the contributing guide
• [] The relevant docs, if any, have been updated or created
• [] The relevant tests, if any, have been updated or created

Screen recording or screenshot

Tests

[ What tests were added or updated to ensure the changes work as expected? ]

---

Summary by cubic

Adds agent skills to the VS Code/JetBrains extension so the model can discover project-specific skills and read detailed instructions via a new tool.

• New Features

  • Load skills from SKILL.md in .continue/skills/ and .claude/skills/ (workspace and global).
  • Parse YAML frontmatter + markdown content.
  • New read_skill tool to fetch a skill by name; includes supporting files and returns a friendly not-found error.

• Migration

  • Add SKILL.md under .continue/skills/ or .claude/skills/ with YAML frontmatter (name, description) and instructions in the body.
  • Reload Continue to pick up skills.

^{Written for commit f5c3452. Summary will update on new commits.}

---

Continue Tasks

Status	Task	Actions
✅ Merged	Update docs on PR	PR · View
▶️ Queued	Optimize Website Performance	View

_{Powered by Continue}

[continue]

Learn more

---

All Green is an AI agent that automatically:

✅ Addresses code review comments

✅ Fixes failing CI checks

✅ Resolves merge conflicts

---

Unsubscribe from All Green comments

[cubic-dev-ai]

4 issues found across 17 files

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="core/config/markdown/loadMarkdownSkills.ts">

<violation number="1" location="core/config/markdown/loadMarkdownSkills.ts:48">
P2: Magic number `slice(7)` is unclear and fragile. This appears to strip the `file://` URI scheme prefix, but should use a named constant or proper URI parsing utility for clarity and safety. Consider extracting this to a helper function or adding a comment explaining the intent.</violation>
</file>

<file name="core/llm/skills.ts">

<violation number="1" location="core/llm/skills.ts:20">
P2: Remove debug `console.log` statement before merging. The `debug1` prefix indicates this is temporary debugging code that should not be in production.</violation>
</file>

<file name="gui/src/redux/util/constructMessages.ts">

<violation number="1" location="gui/src/redux/util/constructMessages.ts:218">
P0: Critical bug: `systemMessageWithSkills` is computed but never used. `finalSystemMessage` is assigned `systemMessageWithRules` instead of `systemMessageWithSkills`, causing the skills feature to have no effect.</violation>

<violation number="2" location="gui/src/redux/util/constructMessages.ts:220">
P0: Same bug: should use `systemMessageWithSkills` instead of `systemMessageWithRules` to include skills when appending conversation summary.</violation>
</file>

_{Reply to cubic to teach it or ask questions. Tag @cubic-dev-ai to re-run a review.}

[continue]

I've created documentation for the agent skills feature in a separate branch: docs-agent-skills

The documentation includes:

• Overview page at docs/customize/skills.mdx
• Detailed guide at docs/customize/deep-dives/skills.mdx
• Updated navigation in docs/docs.json
• Updated reference at docs/reference.mdx

Since your PR is from a fork, I can't create a PR directly against your agent-skills branch. Here are the options:

1. Cherry-pick my commit into your branch:

   git fetch https://github.com/continuedev/continue docs-agent-skills
   git cherry-pick 684d362

2. Merge my branch into yours:

   git remote add continuedev-main https://github.com/continuedev/continue
   git fetch continuedev-main docs-agent-skills
   git merge continuedev-main/docs-agent-skills

3. I can create a separate PR against main that will merge after yours

Let me know which approach you prefer!

[cubic-dev-ai]

No issues found across 13 files

[sestinj]

Let's change where we look for them to be:

• .continue/skills//SKILL.md
• .claude/skills//SKILL.md

[cubic-dev-ai]

11 issues found across 142 files (changes from recent commits).

Note: This PR contains a large number of files. cubic only reviews up to 75 files per PR, so some files may not have been reviewed.

Prompt for AI agents (all issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="docs/guides/sanity-mcp-continue-cookbook.mdx">

<violation number="1">
P2: Broken internal docs link: `/mission-control/configs/intro` does not exist (use the existing Understanding Configs guide instead).</violation>
</file>

<file name="core/tools/implementations/runTerminalCommand.vitest.ts">

<violation number="1">
P2: New vscode-remote URI tests only assert that the call resolves, so the targeted pathname/decoding/file:// preference behaviors are unverified and could regress undetected.</violation>
</file>

<file name="core/tools/implementations/resolveWorkingDirectory.vitest.ts">

<violation number="1">
P2: Test asserts Unix-only file URL result and will fail on Windows CI runners</violation>

<violation number="2">
P2: Test duplicates resolver logic instead of calling production implementation, so changes in real code won’t be covered and tests can give false positives.</violation>
</file>

<file name="docs/guides/cloud-agents/cloud-agents-vs-ci.mdx">

<violation number="1">
P2: Mermaid code block is not closed because the closing ``` is indented by 4 spaces (closing fences can only be indented up to 3), causing subsequent content to render inside the code block.</violation>
</file>

<file name="core/tools/implementations/readSkill.ts">

<violation number="1" location="core/tools/implementations/readSkill.ts:31">
P1: Skill prompt now leaks absolute file URIs for the entire skill directory to the model, exposing user-specific paths and directory listing</violation>
</file>

<file name="docs/agents/create-and-edit.mdx">

<violation number="1">
P2: Broken Supabase card link: href value includes a leading space, producing an invalid URL</violation>
</file>

<file name="docs/guides/cloud-agents/from-task-to-automation.mdx">

<violation number="1">
P2: Stage 1 content is indented 4 spaces, causing the goal and lists to render as a code block instead of normal text.</violation>
</file>

<file name="docs/guides/cloud-agents/when-to-use-cloud-agents.mdx">

<violation number="1">
P3: Missing trailing newline at EOF; Prettier format check for .md files will fail</violation>
</file>

<file name="extensions/cli/package.json">

<violation number="1">
P2: @sentry/profiling-node bumped to 9.47.1 but @sentry/node left at 9.43.0, violating Sentry’s requirement that profiling and main SDK versions match</violation>
</file>

<file name="docs/guides/cloud-agents/automated-security-remediation-with-snyk.mdx">

<violation number="1">
P2: Bullets inside `<Card>` are indented ≥4 spaces, so they render as a code block instead of a list.</violation>
</file>

_{Reply with feedback, questions, or to request a fix. Tag @cubic-dev-ai to re-run a review.}

[sestinj]

@uinstinct do you think we could add support for skills in the CLI as well?