@mcp-s/cli 0.0.11 → 0.0.12
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/SKILL.md +10 -0
- package/dist/SKILL.md +10 -0
- package/dist/daemon.js +1 -1
- package/dist/index.js +1 -1
- package/package.json +1 -1
package/SKILL.md
CHANGED
|
@@ -41,11 +41,21 @@ Every command consumes tokens. Pick the cheapest operation that gets the job don
|
|
|
41
41
|
- **Pipe output** through shell tools (`head`, `grep`, `jq`) to trim large responses before they enter context.
|
|
42
42
|
- Never fetch a full list when the user only needs a few fields (names, links, IDs).
|
|
43
43
|
|
|
44
|
+
### When a filter produces no output
|
|
45
|
+
|
|
46
|
+
When you pipe through a filter and get empty output, **do not** fall back to the raw unfiltered call. Use the **probe-then-filter** pattern:
|
|
47
|
+
|
|
48
|
+
1. **Probe**: fetch a *single* item (e.g., `limit: 1`, or a dedicated get-one tool) without any pipe.
|
|
49
|
+
2. **Learn**: read the response to understand the actual field names, nesting, and types.
|
|
50
|
+
3. **Re-run**: call the bulk tool again with a corrected filter based on what you learned.
|
|
51
|
+
|
|
44
52
|
### Decision heuristic
|
|
45
53
|
|
|
46
54
|
- User mentions a specific tool or domain keyword → `grep` for it.
|
|
47
55
|
- User asks a broad question across all tools → list tools first, then act.
|
|
56
|
+
- Task involves multiple items of the same type → use one bulk/list call, not N individual calls.
|
|
48
57
|
- Tool may return large payloads → use filter params + pipe to trim output.
|
|
58
|
+
- Filter script produced no output → **probe one item** to learn the shape, then re-run with a corrected filter. Never fall back to the raw unfiltered call.
|
|
49
59
|
|
|
50
60
|
## Examples
|
|
51
61
|
|
package/dist/SKILL.md
CHANGED
|
@@ -41,11 +41,21 @@ Every command consumes tokens. Pick the cheapest operation that gets the job don
|
|
|
41
41
|
- **Pipe output** through shell tools (`head`, `grep`, `jq`) to trim large responses before they enter context.
|
|
42
42
|
- Never fetch a full list when the user only needs a few fields (names, links, IDs).
|
|
43
43
|
|
|
44
|
+
### When a filter produces no output
|
|
45
|
+
|
|
46
|
+
When you pipe through a filter and get empty output, **do not** fall back to the raw unfiltered call. Use the **probe-then-filter** pattern:
|
|
47
|
+
|
|
48
|
+
1. **Probe**: fetch a *single* item (e.g., `limit: 1`, or a dedicated get-one tool) without any pipe.
|
|
49
|
+
2. **Learn**: read the response to understand the actual field names, nesting, and types.
|
|
50
|
+
3. **Re-run**: call the bulk tool again with a corrected filter based on what you learned.
|
|
51
|
+
|
|
44
52
|
### Decision heuristic
|
|
45
53
|
|
|
46
54
|
- User mentions a specific tool or domain keyword → `grep` for it.
|
|
47
55
|
- User asks a broad question across all tools → list tools first, then act.
|
|
56
|
+
- Task involves multiple items of the same type → use one bulk/list call, not N individual calls.
|
|
48
57
|
- Tool may return large payloads → use filter params + pipe to trim output.
|
|
58
|
+
- Filter script produced no output → **probe one item** to learn the shape, then re-run with a corrected filter. Never fall back to the raw unfiltered call.
|
|
49
59
|
|
|
50
60
|
## Examples
|
|
51
61
|
|
package/dist/daemon.js
CHANGED
package/dist/index.js
CHANGED