Claude Skill

man

The forge's manual, read from its own definitions — bare = the commands and the working methods, with a command = its purpose, arguments and roster, with a method = its paragraph and skill, with a need in free words = the way the forge has for it

LLM Mart · 0 points · 0 views 0 listing impressions 0 install-command copies
Download pche-broken-artist-forge-of-thought-.claude_skills_man-3ef759b.zip · 1 KB

Install

skills CLI npx skills add https://github.com/pche-broken-artist/forge-of-thought/tree/main/.claude/skills/man
Claude Code claude plugin marketplace add https://llmmart.ai/marketplace.json && claude plugin install pche-broken-artist-forge-of-thought@llmmart
Git git clone https://github.com/pche-broken-artist/forge-of-thought.git

The skills CLI installs just this skill, for any of its supported agents. Claude Code installs the whole pche-broken-artist/forge-of-thought collection as a plugin from our marketplace. Git is the plain clone.

Skill manifest

The manual of the forge (POS.1190): a reader, never a text of its own. Everything it prints is read at the moment of the call from the files that own it — CLAUDE.md, the skills, the agents, the state and genre files — and shortened to the line; nothing is stated here that lives there. Read-only; it runs no command it describes. Alias: /manual (.claude/skills/manual/SKILL.md). The word help is Claude Code's own command, hence the Unix name.

Bare /man — the overview.

  1. Read the Commands table of CLAUDE.md and print it as a list: the command with its arguments, then its purpose cut to one line. /man and /manual included.
  2. Read the Working methods section of CLAUDE.md and print each method's name with the first sentence of its paragraph.
  3. Close with one line: /man <command> for a command's page, /man <method> for a method's.

/man <command> — the page of one command.

  1. Resolve $0 to .claude/skills/<command>/SKILL.md; if none exists, say so and list the commands (step 1 of the overview).
  2. Print the command's purpose (its description), its arguments (its argument-hint, else the CLAUDE.md row) and the CLAUDE.md row's purpose in full.
  3. Print the roster the command dispatches over, where its skill names one: the same scan the bare command makes, read from that skill, each entry with the description of its own file and, where the file carries one, its ## Lens section — for a genre its Elicitation checklist — in full, so that the user knows what a check, a lens, a persona or a genre looks for before running it. /man adds what each entry looks for and runs nothing.
  4. Close with one line naming the skill file, for the reader who wants the whole procedure.

/man <method> — the page of one working method. Resolve $0 against the bold names of the Working methods section of CLAUDE.md (case-insensitive, hyphens and spaces alike); print the paragraph in full and, where it names a skill that holds the method's shape, that skill's path. Unknown, and not a question: say so and list the methods (step 2 of the overview).

/man <question> — the way the forge has for a need said in free words. Anything that resolves to no command and no method is a question (POS.1190).

  1. Read the documentation's index, docs/README.md: every page carries the one question it answers. Choose the pages whose question meets the need. Where the need can be read two ways that lead to different pages, ask once, in one sentence, and wait.
  2. Read the pages chosen; where a page names a skill or a definition and the detail matters, read it too. Nothing is answered from memory: every mechanism the answer names was read in this run.
  3. Answer in plain sentences, in the conversation language: the way the forge has first, with what one types; then what matters to know on that way; where several ways fit, one recommended with its reason, once. Where the way leaves the forge — a repository to initialise, a program to install — say so plainly, since the forge does not do it. The pages read are named by path once, at the end.
  4. Where the forge has nothing, say so; name the open thread of the forge's own project if one carries the need (projects/forge/threads/), and mark what Claude would do without a mechanism as his own (POS.1370).
  5. Run nothing, write nothing, decide nothing: the answer shows a way.

Language: the conversation language (CLAUDE.md, prime directive 6 — translate on write), the whole page, descriptions and Lens sections included; only the notation stays English — command names, file paths, IDs, front-matter keys and the bold names of the methods, each followed by its translation once.

Files (forge-of-thought)
  • SKILL.md 4 KB
    ---
    description: The forge's manual, read from its own definitions — bare = the commands and the working methods, with a command = its purpose, arguments and roster, with a method = its paragraph and skill, with a need in free words = the way the forge has for it
    argument-hint: "[command | method | question]"
    ---
    
    The manual of the forge (POS.1190): a reader, never a text of its
    own. Everything it prints is read at the moment of the call from the
    files that own it — CLAUDE.md, the skills, the agents, the state and
    genre files — and shortened to the line; nothing is stated here that
    lives there. Read-only; it runs no command it describes. Alias:
    `/manual` (`.claude/skills/manual/SKILL.md`). The word `help` is
    Claude Code's own command, hence the Unix name.
    
    **Bare `/man` — the overview.**
    1. Read the Commands table of CLAUDE.md and print it as a list: the
       command with its arguments, then its purpose cut to one line.
       `/man` and `/manual` included.
    2. Read the Working methods section of CLAUDE.md and print each
       method's name with the first sentence of its paragraph.
    3. Close with one line: `/man <command>` for a command's page,
       `/man <method>` for a method's.
    
    **`/man <command>` — the page of one command.**
    1. Resolve `$0` to `.claude/skills/<command>/SKILL.md`; if none
       exists, say so and list the commands (step 1 of the overview).
    2. Print the command's purpose (its `description`), its arguments
       (its `argument-hint`, else the CLAUDE.md row) and the CLAUDE.md
       row's purpose in full.
    3. Print the roster the command dispatches over, where its skill
       names one: the same scan the bare command makes, read from that
       skill, each entry with the `description` of its own file and,
       where the file carries one, its `## Lens` section — for a genre
       its Elicitation checklist — in full, so that the user knows what
       a check, a lens, a persona or a genre looks for before running
       it. `/man` adds what each entry looks for and runs nothing.
    4. Close with one line naming the skill file, for the reader who
       wants the whole procedure.
    
    **`/man <method>` — the page of one working method.** Resolve `$0`
    against the bold names of the Working methods section of CLAUDE.md
    (case-insensitive, hyphens and spaces alike); print the paragraph in
    full and, where it names a skill that holds the method's shape, that
    skill's path. Unknown, and not a question: say so and list the methods
    (step 2 of the overview).
    
    **`/man <question>` — the way the forge has for a need said in free
    words.** Anything that resolves to no command and no method is a
    question (POS.1190).
    1. Read the documentation's index, `docs/README.md`: every page
       carries the one question it answers. Choose the pages whose
       question meets the need. Where the need can be read two ways that
       lead to different pages, ask once, in one sentence, and wait.
    2. Read the pages chosen; where a page names a skill or a definition
       and the detail matters, read it too. Nothing is answered from
       memory: every mechanism the answer names was read in this run.
    3. Answer in plain sentences, in the conversation language: the way
       the forge has first, with what one types; then what matters to
       know on that way; where several ways fit, one recommended with
       its reason, once. Where the way leaves the forge — a repository
       to initialise, a program to install — say so plainly, since the
       forge does not do it. The pages read are named by path once, at
       the end.
    4. Where the forge has nothing, say so; name the open thread of the
       forge's own project if one carries the need
       (`projects/forge/threads/`), and mark what Claude would do without
       a mechanism as his own (POS.1370).
    5. Run nothing, write nothing, decide nothing: the answer shows a
       way.
    
    Language: the conversation language (CLAUDE.md, prime directive 6 —
    translate on write), the whole page, descriptions and Lens sections
    included; only the notation stays English — command names, file
    paths, IDs, front-matter keys and the bold names of the methods,
    each followed by its translation once.
    

Comments (0)

Sign in to join the conversation.

No comments yet.

Reviews (0)

No reviews yet.