You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on May 18, 2026. It is now read-only.
Copy file name to clipboardExpand all lines: .claude-plugin/plugin.json
+1-1Lines changed: 1 addition & 1 deletion
Original file line number
Diff line number
Diff line change
@@ -1,6 +1,6 @@
1
1
{
2
2
"name": "claude-agent-sdk",
3
-
"description": "Build autonomous AI agents with Claude Agent SDK. TypeScript v0.2.123 | Python v0.1.71. API reference for query(), hooks, subagents, MCP, permissions, sandbox, structured outputs, and sessions.",
3
+
"description": "Build autonomous AI agents with Claude Agent SDK. TypeScript v0.2.126 | Python v0.1.72. API reference for query(), hooks, subagents, MCP, permissions, sandbox, structured outputs, and sessions.",
Copy file name to clipboardExpand all lines: README.md
+2-2Lines changed: 2 additions & 2 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -2,7 +2,7 @@
2
2
3
3
A self-updating Claude Code skill for building AI agents with the Claude Agent SDK — covering both [TypeScript](https://github.com/anthropics/claude-agent-sdk-typescript) and [Python](https://github.com/anthropics/claude-agent-sdk-python).
await q.getContextUsage(); // Get context window usage breakdown by category — returns SDKControlGetContextUsageResponse (v0.2.123)
537
+
await q.getContextUsage(); // Get context window usage breakdown by category — returns SDKControlGetContextUsageResponse (v0.2.126)
538
538
await q.readFile(path, { maxBytes?, encoding?: 'utf-8' | 'base64' }); // Read a file from the session filesystem (gated by same read-permission rules as Read tool); returns SDKControlReadFileResponse | null. Use 'base64' for binary files like images.
**Note**:PR [#305](https://github.com/anthropics/claude-agent-sdk-typescript/issues/305) (pending) adds a libc probe to select the correct variant at runtime.
**Symptom**:Setting`skillOverrides: { "<name>": "off" }` (via`settings`optionor`~/.claude/settings.json`) preventstheskillfromrunning, buttheskillstillappearsinthe`<skills>`listinginjectedintothemodel's system prompt every turn. The model is told the skill exists, then gets `"Skill <name> is disabled"` when it tries to invoke it — causing confusing "I know about X but cannot use it" behavior ([#291](https://github.com/anthropics/claude-agent-sdk-typescript/issues/291))
**Impact**:Skillsthatshouldbehiddenfromthemodel's awareness (e.g., internal or dangerous skills) remain visible. The model wastes turns attempting to use disabled skills, and may behave unpredictably when invocation fails.
**Note**:`skills: []`disablestheSDK's own skills context filter but the CLI may still inject built-in skills. Use `skillOverrides: { "*": "off" }` combined with an explicit `skills` allowlist for the strictest control.
Both SDKs received version bumps today: TypeScript v0.2.123 → v0.2.126 and Python v0.1.71 → v0.1.72. The update agent completed successfully in 124s. Research agents ran — the TypeScript agent added Known Issue #55 (`skillOverrides` fails to hide skills from the model's listing), while the Python agent applied only version reference updates. The typecheck step failed on its initial run (stale-comma false-positive in the import-check script, a recurring issue), triggering one mend pass; verification subsequently passed 28/28 checks on attempt 2. All outcomes resolved as success.
No issue state changes or new bug issues detected.
29
+
30
+
## Update Agent
31
+
32
+
Version bumps applied:
33
+
34
+
|| Old | New |
35
+
|---|---|---|
36
+
| TypeScript SDK | v0.2.123 | v0.2.126 |
37
+
| Python SDK | v0.1.71 | v0.1.72 |
38
+
39
+
**Files modified** (11 files):
40
+
-`README.md` — version badges updated
41
+
-`SKILL.md` — version refs updated
42
+
-`SKILL-typescript.md` — version bumped, `SDKAPIRetryMessage` and `getContextUsage()` version tags updated to v0.2.126
43
+
-`SKILL-python.md` — version bumped throughout; `AssistantMessage`, `ResultMessage`, `AgentDefinition` field version tags updated to v0.1.72; custom tools note updated to v0.1.72+; `@tool``Annotated` parameter description updated
44
+
-`rules/claude-agent-sdk-ts.md` — version ref updated to v0.2.126
45
+
-`rules/claude-agent-sdk-py.md` — version ref updated
46
+
-`templates/typescript/package.json` — pinned version updated
47
+
-`templates/python/pyproject.toml` — pinned version updated
48
+
-`.claude-plugin/plugin.json` — version ref updated
49
+
-`agent/state.json` — version fields updated; `lastAuditedVersion` set to v0.2.126 (TS) and lastUpdated refreshed
50
+
-`scripts/check-versions.sh` — version refs updated
51
+
52
+
Verification: passed 28/28 on attempt 2 after 1 mend run. No mending of functional content required — mend fixed the typecheck false-positive only.
53
+
54
+
## Research
55
+
56
+
### TypeScript Research (542s, 80 turns)
57
+
58
+
-**Issues evaluated**: scanned up to issue #291 and related
59
+
-**New Known Issues added**: 1
60
+
61
+
| Issue | Title | Verdict | Reason |
62
+
|-------|-------|---------|--------|
63
+
|[#291](https://github.com/anthropics/claude-agent-sdk-typescript/issues/291)|`skillOverrides: { "name": "off" }` blocks invocation but does not hide skill from model's listing |**added** (KI #55) | Skill loader checks `skillOverrides` at invocation time but not during listing-injection pass; model is told skill exists then gets a "disabled" error — confusing behavior with clear workaround (`skills` allowlist option) |
64
+
65
+
### Python Research (306s, 90 turns)
66
+
67
+
-**Issues evaluated**: recent issues reviewed
68
+
-**New Known Issues added**: 0
69
+
- No new actionable bugs found; all changes were version reference updates cascading from the v0.1.71 → v0.1.72 bump
70
+
71
+
## Errors
72
+
73
+
### Typecheck — Exit Code 1 (recovered by mending)
74
+
75
+
**Last output:**
76
+
```
77
+
OK: query_with_tools.py — valid syntax
78
+
OK: sandbox_config.py — valid syntax
79
+
OK: session_management.py — valid syntax
80
+
OK: structured_output.py — valid syntax
81
+
OK: subagents_orchestration.py — valid syntax
82
+
83
+
=== Python import resolution check ===
84
+
File "<string>", line 1
85
+
from claude_agent_sdk import (,AgentDefinition,ClaudeAgentOptions,...
86
+
^
87
+
SyntaxError: invalid syntax
88
+
FAIL: Some template imports don't resolve
89
+
FAIL: Cannot import: (
90
+
91
+
=========================================
92
+
Passed: 13
93
+
Failed: 2
94
+
=========================================
95
+
96
+
TEMPLATE CHECK FAILED — 2 issue(s)
97
+
```
98
+
99
+
**Likely cause**: Stale leading comma in a generated import list in the check script — a recurring false-positive where the import-check generator produces `(,AgentDefinition,...` instead of `(AgentDefinition,...`. Not a real code defect.
100
+
101
+
**Resolution**: Mend agent fixed the comma in one pass (21 turns, $0.21); verify passed 28/28 on attempt 2.
102
+
103
+
**Suggested fix**: The import-check script should strip leading commas before eval — this false-positive has recurred across multiple days.
![github-actions[bot]](https://avatars.githubusercontent.com/in/15368?v=4&size=40){kind=avatar}
0 commit comments