Skip to content

Commit 0b4d62b

Browse files
SimonSchubertSimonSchubert
committed
Add commands
1 parent 3672d62 commit 0b4d62b

4 files changed

Lines changed: 306 additions & 0 deletions

File tree

‎assets/commands/agenthatch.md‎

Lines changed: 97 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,97 @@
1+
# TAGLINE
2+
3+
Compile markdown skills into standalone Python agents
4+
5+
# TLDR
6+
7+
**Initialize** configuration and the LLM provider
8+
9+
```agenthatch init```
10+
11+
**Register** a skill file
12+
13+
```agenthatch skills add [./my-skill/SKILL.md]```
14+
15+
**Compile** a skill into an agent
16+
17+
```agenthatch hatch [my-skill]```
18+
19+
**Run** a hatched agent interactively
20+
21+
```agenthatch run [my-skill]```
22+
23+
**List** registered skills
24+
25+
```agenthatch skills list```
26+
27+
**Diagnose** the environment
28+
29+
```agenthatch doctor```
30+
31+
# SYNOPSIS
32+
33+
**agenthatch** _command_ [_arguments_] [_options_]
34+
35+
# DESCRIPTION
36+
37+
**agenthatch** is a compiler that turns a markdown **SKILL.md** definition into a standalone, runnable Python AI agent. It parses the skill file, runs an AI inference pass to infer tools and structure, and generates a complete Python package with its own CLI entry point, type-annotated tool definitions, and MCP integration.
38+
39+
Registered skills live in a local index (the "skillhouse"). The typical flow is to register a skill, **hatch** it to generate code, then **run** the resulting agent in an interactive terminal UI.
40+
41+
# COMMANDS
42+
43+
**init**
44+
> Initialize configuration and set up the LLM provider.
45+
46+
**skills add** _path_
47+
> Register a SKILL.md file in the skillhouse index.
48+
49+
**skills list**
50+
> Display all registered skills.
51+
52+
**skills delete** _name_
53+
> Remove a skill from the skillhouse.
54+
55+
**hatch** _name_
56+
> Run the full compilation pipeline (parse, AI inference, code generation).
57+
58+
**run** _name_
59+
> Launch a hatched agent in interactive terminal UI mode.
60+
61+
**search** _query_
62+
> Query the skillhouse index.
63+
64+
**assemble**
65+
> Rebuild an existing skillhouse agent.
66+
67+
**doctor**
68+
> Diagnose the environment and dependencies.
69+
70+
# PARAMETERS
71+
72+
**--no-generate**
73+
> Skip the code-generation phase so the inferred spec can be reviewed first (used with **hatch**).
74+
75+
**--force**
76+
> Overwrite an existing hatched agent.
77+
78+
**--dry-run**
79+
> Preview the output without writing files.
80+
81+
# CAVEATS
82+
83+
Requires Python 3.11 or later. Compilation calls an LLM provider (OpenAI, DeepSeek, Anthropic, or any OpenAI-compatible endpoint), configured via a TOML file, so an API key and network access are needed for the **hatch** step.
84+
85+
# HISTORY
86+
87+
**agenthatch** is written in Python and released under the MIT license. It was published in 2026 around the idea of treating markdown skill files as source code that compiles into deployable agents.
88+
89+
# SEE ALSO
90+
91+
[python](/man/python)(1), [pipx](/man/pipx)(1), [claude](/man/claude)(1)
92+
93+
# RESOURCES
94+
95+
```[Source code](https://github.com/agenthatch/agenthatch)```
96+
97+
<!-- verified: 2026-06-19 -->

‎assets/commands/crawlie.md‎

Lines changed: 111 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,111 @@
1+
# TAGLINE
2+
3+
Technical SEO and GEO crawler for the terminal
4+
5+
# TLDR
6+
7+
**Crawl** an entire site and print a report
8+
9+
```crawlie crawl [https://example.com] --format pretty```
10+
11+
**Audit** a single page
12+
13+
```crawlie audit [https://example.com/pricing]```
14+
15+
**Save** an HTML report to a file
16+
17+
```crawlie crawl [https://example.com] --format html -o [report.html]```
18+
19+
**Limit** the number of pages fetched
20+
21+
```crawlie crawl [https://example.com] --max-pages [100]```
22+
23+
**Fail** a CI build when errors are found
24+
25+
```crawlie crawl [https://example.com] --fail-on error```
26+
27+
**Explain** why a finding matters
28+
29+
```crawlie explain [geo-not-answerable]```
30+
31+
# SYNOPSIS
32+
33+
**crawlie** _subcommand_ [_url_] [_options_]
34+
35+
# DESCRIPTION
36+
37+
**crawlie** is a fast, open-source crawler that audits a website for technical SEO and GEO (Generative Engine Optimization) issues. It seeds from **sitemap.xml**, respects **robots.txt**, follows internal links, and applies a set of rules covering broken links, missing or duplicate metadata, performance hints, mobile and international optimization, and AI-search readiness.
38+
39+
Results can be printed to the terminal or exported as JSON, CSV, or a shareable HTML report. The same engine is exposed as an MCP server (**crawlie-mcp**) for use inside agent workflows.
40+
41+
# SUBCOMMANDS
42+
43+
**crawl** _url_
44+
> Crawl an entire site, seeding from the sitemap and following internal links.
45+
46+
**audit** _url_
47+
> Audit a single page or a specific set of pages.
48+
49+
**explain** _finding_
50+
> Describe why a finding matters and how to fix it.
51+
52+
**reports**
53+
> List the local report history.
54+
55+
**report** _id_
56+
> Retrieve a specific saved report by its ID.
57+
58+
# PARAMETERS
59+
60+
**--format** _fmt_
61+
> Output format: **pretty** (terminal), **json**, **csv**, or **html** (default: json).
62+
63+
**-o**, **--output** _file_
64+
> Write the report to a file instead of standard output.
65+
66+
**--max-pages** _n_
67+
> Maximum number of pages to fetch (default: 500).
68+
69+
**--max-depth** _n_
70+
> Maximum click depth from the seed URL.
71+
72+
**--concurrency** _n_
73+
> Number of parallel requests (default: 16).
74+
75+
**--include** _glob_
76+
> Only crawl URLs matching the pattern.
77+
78+
**--exclude** _glob_
79+
> Skip URLs matching the pattern.
80+
81+
**--no-robots**
82+
> Do not parse or honor robots.txt.
83+
84+
**--no-sitemap**
85+
> Do not seed the crawl from sitemap.xml.
86+
87+
**--no-external**
88+
> Skip checking external links.
89+
90+
**--severity** _level_
91+
> Filter findings by severity: **error**, **warning**, or **notice**.
92+
93+
**--save**
94+
> Store the result in the local report history.
95+
96+
**--fail-on** _level_
97+
> Exit with a non-zero status when findings of **error** or **warning** severity are present (useful in CI).
98+
99+
# HISTORY
100+
101+
**crawlie** is written in Rust (with a TypeScript desktop companion) and released under the MIT license. It was published in 2026 as a free, agent-friendly alternative to commercial SEO crawlers, adding GEO checks aimed at how AI search engines read a site.
102+
103+
# SEE ALSO
104+
105+
[wget](/man/wget)(1), [curl](/man/curl)(1), [httrack](/man/httrack)(1), [scrapy](/man/scrapy)(1)
106+
107+
# RESOURCES
108+
109+
```[Source code](https://github.com/spronta/crawlie)```
110+
111+
<!-- verified: 2026-06-19 -->

‎assets/commands/index.txt‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -120,6 +120,7 @@ age.md
120120
agent-skills-eval.md
121121
agent-vault.md
122122
agentbox.md
123+
agenthatch.md
123124
agetty.md
124125
agg.md
125126
agy.md
@@ -1403,6 +1404,7 @@ crane-version.md
14031404
crane.md
14041405
crates-tui.md
14051406
crawl.md
1407+
crawlie.md
14061408
crc32.md
14071409
crda.md
14081410
create-image.md
@@ -6508,6 +6510,7 @@ rehash.md
65086510
reindexdb.md
65096511
rekal.md
65106512
rekor-cli.md
6513+
relaymux.md
65116514
relic.md
65126515
remind.md
65136516
remmina.md

‎assets/commands/relaymux.md‎

Lines changed: 95 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,95 @@
1+
# TAGLINE
2+
3+
Local tmux-based meta-harness for coding agents
4+
5+
# TLDR
6+
7+
**Set up** relaymux with Telegram remote control
8+
9+
```relaymux setup --telegram --telegram-bot-token '[token]'```
10+
11+
**Check** service status
12+
13+
```relaymux status```
14+
15+
**Launch** an agent on a repository
16+
17+
```relaymux launch --repo [~/code/my-app] --agent [claude] --name [fix-tests] --prompt "[Fix the failing tests]"```
18+
19+
**Send** a test notification
20+
21+
```relaymux notify --from [test] --reply-mode telegram --message "[hello]"```
22+
23+
# SYNOPSIS
24+
25+
**relaymux** _subcommand_ [_options_]
26+
27+
# DESCRIPTION
28+
29+
**relaymux** is a lightweight local meta-harness that runs command-line coding agents inside visible **tmux** sessions while letting you monitor and reply to them remotely. Each agent run gets its own tmux tab on your machine, so the work stays local and inspectable, and Telegram acts as the remote control channel for status updates and prompts.
30+
31+
It orchestrates existing agent CLIs (for example **claude**, **codex**, or **pi**) rather than replacing them, making it useful for kicking off and supervising long-running agent tasks from a phone.
32+
33+
# SUBCOMMANDS
34+
35+
**setup**
36+
> Initialize configuration, including Telegram integration.
37+
38+
**status**
39+
> Report the status of the relaymux service and running agents.
40+
41+
**launch**
42+
> Start an agent run on a given repository with a prompt.
43+
44+
**notify**
45+
> Send a manual notification through the configured reply channel.
46+
47+
**db**
48+
> Perform database operations (requires the system **sqlite3** CLI).
49+
50+
# PARAMETERS
51+
52+
**--telegram**
53+
> Enable Telegram integration (used with **setup**).
54+
55+
**--telegram-bot-token** _token_
56+
> Telegram bot authentication token.
57+
58+
**--repo** _path_
59+
> Repository or workspace path for the agent run.
60+
61+
**--agent** _name_
62+
> Agent CLI to launch (for example **claude**, **codex**, or **pi**).
63+
64+
**--name** _name_
65+
> Label for the agent run.
66+
67+
**--prompt** _text_
68+
> Task instructions passed to the agent.
69+
70+
**--from** _source_
71+
> Notification source label (used with **notify**).
72+
73+
**--reply-mode** _channel_
74+
> Reply channel, for example **telegram**.
75+
76+
**--message** _text_
77+
> Notification body (used with **notify**).
78+
79+
# CAVEATS
80+
81+
Requires Node.js 20+, npm, **tmux**, and a local agent CLI to drive. Database operations need the system **sqlite3** command on the PATH.
82+
83+
# HISTORY
84+
85+
**relaymux** is written in TypeScript and released under the MIT license. It was published in 2026 as a way to supervise local coding agents remotely without handing control to a hosted service.
86+
87+
# SEE ALSO
88+
89+
[tmux](/man/tmux)(1), [screen](/man/screen)(1), [zellij](/man/zellij)(1), [claude](/man/claude)(1)
90+
91+
# RESOURCES
92+
93+
```[Source code](https://github.com/mupt-ai/relaymux)```
94+
95+
<!-- verified: 2026-06-19 -->

0 commit comments

Comments
 (0)