coding_skill

Name: coding_skill
Author: Negai-ai

341stars33forks

How to install coding_skill?

negai-ai/agentclaw/coding_skill

$npx -y skills add Negai-ai/AgentClaw --skill coding_skill

Installs into the current project.

Summary

Inspect, edit, test, refactor, and validate project code inside the configured project directory. Use when the task involves source search, code reading, bug fixes, implementation changes, syntax checks, automated tests, file moves, or refactors.

Files · 1

View on GitHub

SKILL.md

1	# Coding Skill (Primary Playbook)
2
3	This `SKILL.md` is the default guide for common coding tasks.
4	Focus on reliable tool usage and verifiable code changes.
5	Load `references/*` only when needed.
6
7	## 1) Scope and Boundary
8
9	- This skill covers generic coding work: discover, read, edit, refactor, and validate code.
10	- This skill is tool-centric. Keep guidance about product/domain-specific implementation out of the main flow.
11	- All coding-tool paths are resolved under `project_dir`. Paths outside `project_dir` are invalid.
12	- Use minimal edits first; escalate to broad rewrites only when partial edits are unstable.
13	- Do not introduce hardcoded secrets, machine-specific absolute paths, or environment-specific install commands in generated code.
14
15	## 2) Zero-Context Execution Flow (Default)
16
17	When repository context is limited, use this fixed sequence:
18
19	1. `search_code` to locate files/snippets.
20	2. Run quick environment preflight for path/tool consistency:
21	- confirm `project_dir`-relative target paths
22	- avoid bare `pip`; prefer interpreter-scoped install commands when needed
23	3. `read_code` to load exact edit context.
24	4. Apply smallest safe edit (`replace_in_file` or `update_code`).
25	5. For Python edits:
26	- If you used `write_code`: syntax is already validated (pre-write check). Do NOT call `syntax_check` in the same tool call — the file won't exist yet when parallel tools run. Only use `syntax_check` separately for `replace_in_file` / `update_code` edits.
27	- `python -m py_compile <changed_file>.py` (optional, for import-level validation; on Windows you may also use `py -3 -m py_compile`)
28	6. If syntax fails, patch the failing region with local edits first (`replace_in_file` / `update_code`), then re-run checks.
29	7. Re-read changed region (`read_code`) when ambiguity/count errors occurred.
30	8. Report changed files, validations, and unresolved risks.
31
32	Prefer not to skip to later steps when an earlier step fails.
33
34	## 3) Tool Contract (Default)
35
36	### 3.1 `search_code`
37
38	Purpose: text/regex discovery in project files. Returns `file:line:col: matched_text` for each match.
39
40	Important: This tool only returns the matching line text, NOT full function/class definitions or surrounding code. To read the full implementation of a matched symbol, use `read_code` with the file path and line range from search results.
41
42	Required:
43	- `query` (string)
44
45	Common optional:
46	- `path` (default `"."`)
47	- `file_glob` (default `"*/"`)
48	- `use_regex` (default `false`)
49	- `max_results` (default `200`, range `1-1000`)
50
51	Example:
52
53	```json
54	{
55	"query": "class Workflow",
56	"path": "agentclaw",
57	"file_glob": "*/.py"
58	}
59	```
60
61	### 3.2 `read_code`
62
63	Purpose: read file content by line range or Python symbol.
64
65	Required:
66	- `path`
67
68	Optional:
69	- `start_line`, `end_line`
70	- `symbol` (Python only)
71	- `symbol_type` (`auto\|function\|class`)
72
73	Examples:
74
75	```json
76	{
77	"path": "agentclaw/graph/workflow.py",
78	"start_line": 60,
79	"end_line": 180
80	}
81	```
82
83	```json
84	{
85	"path": "agentclaw/node/llm.py",
86	"symbol": "_build_messages",
87	"symbol_type": "function"
88	}
89	```
90
91	### 3.3 `syntax_check`
92
93	Purpose: language-aware syntax diagnostics.
94
95	Required:
96	- `path`
97
98	Optional:
99	- `language` override (`py\|js\|ts\|json\|yaml\|toml`)
100	- `context_lines` (default `3`, range `0-8`)
101	- `include_source_context` (default `true`)
102
103	Example:
104
105	```json
106	{"path":"agentclaw/node/llm.py"}
107	```
108
109	When to use:
110	- immediately after editing Python/JS/TS/JSON/YAML/TOML files
111	- before `py_compile`/register/runtime checks
112
113	When NOT to use:
114	- for semantic/runtime logic debugging (it is syntax-focused)
115	- as a replacement for import/runtime validation
116
117	Common failure shapes -> next action:
118	- `path_not_found` / `Not a file`
119	- align path with `project_dir`, then retry same check
120	- `PY_SYNTAX_ERROR` / syntax diagnostics with line+col
121	- patch failing region via `replace_in_file` or `update_code`, then re-run `syntax_check`
122	- checker unavailable (`node`, `tsc`, `PyYAML`, `tomli`)
123	- report dependency gap and continue with partial validation
124
125	Output format:
126	- Human-readable block first (error message, source context with `>` marker, smart HINT), followed by structured JSON.
127	- HINT line detects common root causes:
128	- `HTML entities (" > ')` → use `write_code` instead of `write_file` to auto-sanitize.
129	- `double-escaped sequences (\\\\n)` → use literal multi-line content or single `\\n`.
130	- `ok=true` + `status=success`: syntax pass with no flagged risk pattern.
131	- `ok=true` + `status=warning`: syntax pass but `risk_diagnostics` is non-empty.
132	- treat as unresolved risk and patch before claiming completion.
133	- `ok=false` + `diagnostics`: syntax failure; repair same region first.
134
135	### 3.4 `update_code`
136
137	Purpose: regex-based code update wi

People who liked this also viewed

6 picks

Type

coding_skill

341stars33forks

View on GitHub

How to install coding_skill?

negai-ai/agentclaw/coding_skill

$npx -y skills add Negai-ai/AgentClaw --skill coding_skill

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Coding Skill (Primary Playbook)
2
3	This `SKILL.md` is the default guide for common coding tasks.
4	Focus on reliable tool usage and verifiable code changes.
5	Load `references/*` only when needed.
6
7	## 1) Scope and Boundary
8
9	- This skill covers generic coding work: discover, read, edit, refactor, and validate code.
10	- This skill is tool-centric. Keep guidance about product/domain-specific implementation out of the main flow.
11	- All coding-tool paths are resolved under `project_dir`. Paths outside `project_dir` are invalid.
12	- Use minimal edits first; escalate to broad rewrites only when partial edits are unstable.
13	- Do not introduce hardcoded secrets, machine-specific absolute paths, or environment-specific install commands in generated code.
14
15	## 2) Zero-Context Execution Flow (Default)
16
17	When repository context is limited, use this fixed sequence:
18
19	1. `search_code` to locate files/snippets.
20	2. Run quick environment preflight for path/tool consistency:
21	- confirm `project_dir`-relative target paths
22	- avoid bare `pip`; prefer interpreter-scoped install commands when needed
23	3. `read_code` to load exact edit context.
24	4. Apply smallest safe edit (`replace_in_file` or `update_code`).
25	5. For Python edits:
26	- If you used `write_code`: syntax is already validated (pre-write check). Do NOT call `syntax_check` in the same tool call — the file won't exist yet when parallel tools run. Only use `syntax_check` separately for `replace_in_file` / `update_code` edits.
27	- `python -m py_compile <changed_file>.py` (optional, for import-level validation; on Windows you may also use `py -3 -m py_compile`)
28	6. If syntax fails, patch the failing region with local edits first (`replace_in_file` / `update_code`), then re-run checks.
29	7. Re-read changed region (`read_code`) when ambiguity/count errors occurred.
30	8. Report changed files, validations, and unresolved risks.
31
32	Prefer not to skip to later steps when an earlier step fails.
33
34	## 3) Tool Contract (Default)
35
36	### 3.1 `search_code`
37
38	Purpose: text/regex discovery in project files. Returns `file:line:col: matched_text` for each match.
39
40	Important: This tool only returns the matching line text, NOT full function/class definitions or surrounding code. To read the full implementation of a matched symbol, use `read_code` with the file path and line range from search results.
41
42	Required:
43	- `query` (string)
44
45	Common optional:
46	- `path` (default `"."`)
47	- `file_glob` (default `"*/"`)
48	- `use_regex` (default `false`)
49	- `max_results` (default `200`, range `1-1000`)
50
51	Example:
52
53	```json
54	{
55	"query": "class Workflow",
56	"path": "agentclaw",
57	"file_glob": "*/.py"
58	}
59	```
60
61	### 3.2 `read_code`
62
63	Purpose: read file content by line range or Python symbol.
64
65	Required:
66	- `path`
67
68	Optional:
69	- `start_line`, `end_line`
70	- `symbol` (Python only)
71	- `symbol_type` (`auto\|function\|class`)
72
73	Examples:
74
75	```json
76	{
77	"path": "agentclaw/graph/workflow.py",
78	"start_line": 60,
79	"end_line": 180
80	}
81	```
82
83	```json
84	{
85	"path": "agentclaw/node/llm.py",
86	"symbol": "_build_messages",
87	"symbol_type": "function"
88	}
89	```
90
91	### 3.3 `syntax_check`
92
93	Purpose: language-aware syntax diagnostics.
94
95	Required:
96	- `path`
97
98	Optional:
99	- `language` override (`py\|js\|ts\|json\|yaml\|toml`)
100	- `context_lines` (default `3`, range `0-8`)
101	- `include_source_context` (default `true`)
102
103	Example:
104
105	```json
106	{"path":"agentclaw/node/llm.py"}
107	```
108
109	When to use:
110	- immediately after editing Python/JS/TS/JSON/YAML/TOML files
111	- before `py_compile`/register/runtime checks
112
113	When NOT to use:
114	- for semantic/runtime logic debugging (it is syntax-focused)
115	- as a replacement for import/runtime validation
116
117	Common failure shapes -> next action:
118	- `path_not_found` / `Not a file`
119	- align path with `project_dir`, then retry same check
120	- `PY_SYNTAX_ERROR` / syntax diagnostics with line+col
121	- patch failing region via `replace_in_file` or `update_code`, then re-run `syntax_check`
122	- checker unavailable (`node`, `tsc`, `PyYAML`, `tomli`)
123	- report dependency gap and continue with partial validation
124
125	Output format:
126	- Human-readable block first (error message, source context with `>` marker, smart HINT), followed by structured JSON.
127	- HINT line detects common root causes:
128	- `HTML entities (" > ')` → use `write_code` instead of `write_file` to auto-sanitize.
129	- `double-escaped sequences (\\\\n)` → use literal multi-line content or single `\\n`.
130	- `ok=true` + `status=success`: syntax pass with no flagged risk pattern.
131	- `ok=true` + `status=warning`: syntax pass but `risk_diagnostics` is non-empty.
132	- treat as unresolved risk and patch before claiming completion.
133	- `ok=false` + `diagnostics`: syntax failure; repair same region first.
134
135	### 3.4 `update_code`
136
137	Purpose: regex-based code update wi

People who liked this also viewed

6 picks

Type