[shhdwi]

Claude Code loads repo skills on its own, so Trail never heard about those runs. On the Skills page, a skill the team uses every day said "Not used yet".

What changes

• The tool-savings PostToolUse hook now also matches Skill. When the agent loads a skill, the hook queues a skill use.
• The prompt hook does the same when the person types /name.
• trail init updates the matcher on machines that already have Trail installed.
• Uses go through the existing activity outbox as a new kind, skill. They're sent with the next sync and retried on failure, and nothing is sent unless the machine is connected.

What counts as a use

• Only skills the repo keeps at .claude/skills/<name>/SKILL.md. Personal and plugin (plugin:name) skills don't count.
• The same skill twice in one session within 2 minutes is one use, so typing /name and the agent then loading it isn't counted twice.
• A use is only sent once this machine has uploaded that skill to the workspace. Until then it's dropped at send time.
• Most tool calls and prompts aren't skill uses, and they return before reading any file.

Privacy: only the skill's name is sent. Never the task, the prompt or the arguments; the tests check this.

The server side is trailhq/trail-cloud#31 (merged).

Tested

• test/trail-skill-use.test.ts: parsing, queueing once per run, only repo skills in workspace repos, and the send-only-after-sync rule.
• Full suite: 3999/4009 pass. The 2 failures are the known graft-to-trail tests that only fail on this machine.
• End to end on a sandbox laptop:
  • The real ~/.claude hook shim fired on a Skill PostToolUse and on a /cut-a-release 4.3 for the Acme invoice fix prompt.
  • The outbox held the name only, with no args.
  • _trail-sync sent both uses, and the local server stored 2 coding-agent runs.
  • The skill page reads "Used 2×".

🤖 Generated with Claude Code

[trailhq-graft]

🌱 graft blast radius

2 areas changed → 3 areas can be affected. 8 dependent symbols, depth 2.
Tests: 2 areas updated their tests.
Tag: @anirudhkumar-nanonets — 5 of 5 areas · Rushabh Nagda — Workspace Activity Sync, Pair Session Setup

flowchart TB
  A0(("Pair Session Setup<br/>4 symbols"))
  A1(("CLI Entry Point<br/>3 symbols"))
  A2(("Trail Initialization<br/>1 symbol"))
  classDef reached fill:#D9EDF3,stroke:#3AA7C9,stroke-width:1.5px,color:#0E313C;
  class A0,A1,A2 reached;

Loading

Can be affected	Symbols	Nearest hop	Reached from
Pair Session Setup	4	src/cloud/lookup-activity.ts:L28-L57 recordLookup — calls, depth 2	Workspace Activity Sync
CLI Entry Point	3	src/cli.ts:L2139-L2213 workspaceCommand — calls, depth 2	Workspace Activity Sync
Trail Initialization	1	src/claude/init.ts:L194-L245 runTrailInit — calls, depth 2	Prompt Hook Processing

Who knows this code — 2 people across 5 areas

Area	Who knows it
Workspace Activity Sync · changed	Rushabh Nagda — 2 commits, last 2d ago · @anirudhkumar-nanonets — 1 commit, last 10d ago
Prompt Hook Processing · changed	@anirudhkumar-nanonets — 22 commits, last yesterday
Pair Session Setup · affected	Rushabh Nagda — 1 commit, last 2d ago · @anirudhkumar-nanonets — 1 commit, last 10d ago
CLI Entry Point · affected	@anirudhkumar-nanonets — 61 commits, last yesterday
Trail Initialization · affected	@anirudhkumar-nanonets — 11 commits, last 8d ago

Ownership is git history over each area's own files, weighted towards recent work (120-day half-life). Merge commits and bots are dropped, and you are dropped from your own PR. A name with no @ has no GitHub handle in its commit email — tag them by hand, or add a .mailmap entry. A suggestion from history, not a CODEOWNERS rule.

All 8 dependent symbols, grouped by area

Pair Session Setup — 4 symbols in 4 files

• src/cloud/lookup-activity.ts:L28-L57 — recordLookup (calls, depth 2)
  28: export function recordLookup(r: AskResult, repo: string, cloud: Cloud | null = readCloud(), savedTokens?: number): void {
• src/cloud/nudge.ts:L269-L323 — waitForConnect (calls, depth 2)
  285: const repos = pickerRepos(pending.repo, env);
• src/cloud/other-repos.ts:L102-L135 — setUpPicked (calls, depth 2)
• src/cloud/pair.ts:L263-L321 — finishPair (calls, depth 2)
  263: export async function finishPair(state: string, repo: string, opts: FinishOptions = {}): Promise<"uploaded" | "expired" | "timed_out" | "failed"> {

CLI Entry Point — 3 symbols in 1 file

• src/cli.ts:L2139-L2213 — workspaceCommand (calls, depth 2)
  2143: .option("--json", "output as JSON")
• src/cli.ts:L2378-L2427 — runTeamCheck (calls, depth 2)
  2378: async function runTeamCheck(dir: string, opts: { base?: string; json?: boolean }): Promise<void> {
• src/cli.ts:L1-L2588 — cli.ts (calls, depth 2)
  5: * graft/ and a clone has the graph. A workspace parent (≥2 git children)

Trail Initialization — 1 symbol in 1 file

• src/claude/init.ts:L194-L245 — runTrailInit (calls, depth 2)
  224: else warnings.push(notPlainJson('.claude/settings.local.json', "trail's status line and allowlist"));

Test signal per changed area — 2 ✓

Reached = a node under a test path has a resolved edge into the changed symbol. It undercounts anything called indirectly — through a CLI, a spawned process or a dynamic import — so read a low ratio as “look here”, never as a coverage gate.

• ✓ Workspace Activity Sync — 1 of 6 reached · 1 test file changed here: test/trail-skill-use.test.ts
  • not reached: repoSkillName, rememberSent, readSent, onlySent, activityEvent
• ✓ Prompt Hook Processing — 4 of 6 reached · 1 test file changed here: test/trail-skill-use.test.ts
  • not reached: main, hookBlocks

8 test suites also reference this code

8 symbols, kept out of the diagram and the table so they cannot crowd out the areas a reviewer has to look at.

• test/claude-hook-timeouts.test.ts
• test/claude-pre-bash-wiring.test.ts
• test/trail-cloud.test.ts
• test/trail-connect.test.ts
• test/trail-init-layout.test.ts
• test/trail-prompt-activity.test.ts
• test/trail-upload-names.test.ts
• test/trail-upload-safety.test.ts

_{graft blast · origin/main...HEAD · depth 2 · 10 changed files}

Open the interactive graph → — click an area to see its dependent symbols at file:line.