claude-session-skill 1.1.4 → 1.1.6
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.
- package/README.md +17 -0
- package/SKILL.md +8 -2
- package/package.json +1 -1
package/README.md
CHANGED
|
@@ -2,6 +2,13 @@
|
|
|
2
2
|
|
|
3
3
|
A [Claude Code](https://docs.anthropic.com/en/docs/claude-code) skill for searching, browsing, and naming past sessions. Indexes all session history and generates AI summaries so you can find any session by keyword, project, or name.
|
|
4
4
|
|
|
5
|
+
This package ships two ways to use the same session index:
|
|
6
|
+
|
|
7
|
+
- **Claude Code skill** — install via `git clone`, use `/session` commands directly in Claude Code
|
|
8
|
+
- **MCP server** — install via `npm`, connect any MCP-compatible client (Claude Desktop, Cursor, etc.)
|
|
9
|
+
|
|
10
|
+
Both read from the same `~/.claude/` session data. Use whichever fits your workflow.
|
|
11
|
+
|
|
5
12
|
## Requirements
|
|
6
13
|
|
|
7
14
|
- [Bun](https://bun.sh) runtime
|
|
@@ -73,6 +80,16 @@ Names are 1–50 characters. Named sessions rank highest in search results. Clea
|
|
|
73
80
|
| Within 24 hours | 1.5× recency boost |
|
|
74
81
|
| Within 7 days | 1.2× recency boost |
|
|
75
82
|
|
|
83
|
+
## Resuming sessions
|
|
84
|
+
|
|
85
|
+
Every session view (list, search, detail) prints a ready-to-run resume command:
|
|
86
|
+
|
|
87
|
+
```
|
|
88
|
+
Resume: cd /path/to/project && claude --resume <session-id>
|
|
89
|
+
```
|
|
90
|
+
|
|
91
|
+
**Why the `cd`?** `claude --resume` is directory-scoped — it only searches for sessions stored under the project folder matching your *current* working directory. Running it from the wrong directory returns `No conversation found` even with a valid session ID. Always run the full `cd ... && claude --resume ...` command as printed.
|
|
92
|
+
|
|
76
93
|
## MCP Server
|
|
77
94
|
|
|
78
95
|
`claude-session-skill` ships a Model Context Protocol (MCP) server so any MCP-compatible client (Claude Desktop, Cursor, etc.) can search sessions without using the CLI skill.
|
package/SKILL.md
CHANGED
|
@@ -40,8 +40,14 @@ Claude Code already saves every user message to `~/.claude/history.jsonl` with s
|
|
|
40
40
|
Entry point: `bun run ~/.claude/skills/session/session.ts <command> [args]`
|
|
41
41
|
|
|
42
42
|
**CRITICAL OUTPUT RULES — read before running anything:**
|
|
43
|
-
1. **Print the FULL command output verbatim
|
|
44
|
-
|
|
43
|
+
1. **Print the FULL command output verbatim, inside a fenced code block.** Wrap ALL output in triple backticks so it renders exactly as-is. Like this:
|
|
44
|
+
````
|
|
45
|
+
```
|
|
46
|
+
<raw output here>
|
|
47
|
+
```
|
|
48
|
+
````
|
|
49
|
+
Do NOT summarize, paraphrase, truncate, or reformat it. Do NOT shorten session IDs — they are full UUIDs and must be shown in full. The skill handles all formatting. Your only job is to relay what it prints, nothing more.
|
|
50
|
+
2. **Do NOT add commentary** after the output. No "What do you need?", no "Here are your sessions:", no interpretation. Just the code block.
|
|
45
51
|
3. **Do NOT ask follow-up questions** after `list` or `search`. The output is self-contained.
|
|
46
52
|
4. **If the user references an already-shown list** (e.g., "the most recent", "the third one", "that one") — use the IDs already printed in the previous output. Do NOT re-run `list` or `show` unnecessarily.
|
|
47
53
|
|
package/package.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
|
1
1
|
{
|
|
2
2
|
"name": "claude-session-skill",
|
|
3
|
-
"version": "1.1.
|
|
3
|
+
"version": "1.1.6",
|
|
4
4
|
"description": "Search, browse, and resume past Claude Code sessions with AI-generated summaries — plus an MCP server for any MCP-compatible client",
|
|
5
5
|
"author": "ITeachYouAI <engineering@iteachyouai.com>",
|
|
6
6
|
"license": "MIT",
|