@olegkoval/agent-skills 1.15.0 → 1.16.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "olko-agent-skills",
3
3
  "description": "Agent-agnostic skill catalog for Codex, Claude, Cursor, and other skill-aware tools.",
4
- "version": "1.14.1",
4
+ "version": "1.15.0",
5
5
  "author": {
6
6
  "name": "Oleg Koval"
7
7
  },
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@olegkoval/agent-skills",
3
- "version": "1.15.0",
3
+ "version": "1.16.0",
4
4
  "private": false,
5
5
  "publishConfig": {
6
6
  "access": "public"
@@ -0,0 +1,116 @@
1
+ ---
2
+ name: relay
3
+ description: >
4
+ Use claude-relay to run a long or rate-limit-prone task autonomously across subscription accounts.
5
+ Invoke this skill whenever the user says "use relay", "run this with relay", "do this overnight",
6
+ "run this autonomously", "use relay for X", or has a multi-hour / multi-step task they want to
7
+ hand off completely. Also use proactively when a task looks like it will take many turns and the
8
+ user hasn't thought about rate limits yet. Guides goal writing, picks the right flags, and starts
9
+ the run.
10
+ license: MIT
11
+ allowed-tools: Bash, Read, Write
12
+ metadata:
13
+ author: Oleg Koval
14
+ tags:
15
+ - relay
16
+ - autonomous
17
+ - rate-limits
18
+ - claude-relay
19
+ ---
20
+
21
+ # relay
22
+
23
+ Runs a task **unattended** using `claude-relay`: drives `claude -p` in a headless loop, writes a
24
+ handoff summary when one account's limit approaches, and switches to the next account automatically.
25
+ You watch logs; relay handles the rest.
26
+
27
+ Binary: `claude-relay` (alias `relay` if added to shell). Config: `~/.claude-relay/relay.config.json`.
28
+
29
+ ---
30
+
31
+ ## Step 1 — Decide: relay or inline?
32
+
33
+ **Use relay when:**
34
+ - Task will take many turns or hours (big refactor, migration, research + implementation)
35
+ - The user wants to hand it off and walk away
36
+ - Rate limits are a real concern (long-running or token-heavy)
37
+
38
+ **Stay inline when:**
39
+ - Task is quick (< ~10 turns)
40
+ - User wants to review each step
41
+ - Task needs interactive input that relay can't provide
42
+
43
+ ---
44
+
45
+ ## Step 2 — Write the goal
46
+
47
+ A relay goal is a self-contained spec a fresh Claude agent can execute cold. Write it to
48
+ `/tmp/relay-goal-<short-name>.md`. It must include:
49
+
50
+ - **What to do** — concrete deliverables, not vague intent
51
+ - **How to verify** — the exact command proving success (`npm test`, `curl /api/health`, etc.)
52
+ - **Scope** — explicit list of what NOT to do (no commits, no PRs, no migrations — unless stated)
53
+ - **Key files** — paths the agent should read first if the context isn't obvious
54
+
55
+ **Template:**
56
+ ```markdown
57
+ ## Goal
58
+ <1-3 sentences: the deliverable>
59
+
60
+ ## Success criteria
61
+ Run `<command>` and confirm it exits 0 / returns the expected result.
62
+
63
+ ## Scope
64
+ - DO: <explicit list>
65
+ - DO NOT: open PRs / push commits / run migrations / touch unrelated files
66
+
67
+ ## Key files
68
+ - <path> — <why it matters>
69
+ ```
70
+
71
+ Good goals are specific. "Add input validation to all POST routes in `app/api/`" beats "improve the API".
72
+
73
+ ---
74
+
75
+ ## Step 3 — Pick flags
76
+
77
+ | Situation | Flag |
78
+ |-----------|------|
79
+ | Task needs uncommitted changes, local `.env`, or a local database | `--in-place` |
80
+ | Everything is committed; task is self-contained | *(default — isolated git worktree)* |
81
+
82
+ When in doubt: if the repo has a `.env` or `node_modules` the task depends on, use `--in-place`.
83
+
84
+ ---
85
+
86
+ ## Step 4 — Run
87
+
88
+ ```bash
89
+ claude-relay run /tmp/relay-goal-<name>.md <absolute-path-to-repo> [--in-place]
90
+ ```
91
+
92
+ Relay logs stream to stderr. The run ends when the agent outputs `RELAY_TASK_COMPLETE` on its own
93
+ line, or hits a hard cap (400 turns / 48h wall clock).
94
+
95
+ ---
96
+
97
+ ## Step 5 — Tell the user
98
+
99
+ After starting, report:
100
+ - The exact command you ran
101
+ - What to watch for: `RELAY_TASK_COMPLETE` in logs, or `claude-relay status --config ~/.claude-relay/relay.config.json`
102
+ - The flag you picked and why (especially if you chose `--in-place`)
103
+
104
+ ---
105
+
106
+ ## Health check
107
+
108
+ If anything seems misconfigured:
109
+ ```bash
110
+ claude-relay doctor --config ~/.claude-relay/relay.config.json
111
+ ```
112
+
113
+ Should show ✓ for both account slots. If a slot shows ✗, tell the user to run:
114
+ ```bash
115
+ claude-relay login <email> # must have claude.ai open in browser as that account first
116
+ ```