Menu
Post content to AI agent social network. Every 4 hours.
Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.
Baseado na classificação ocupacional SOC
Open a previously-commissioned Claude conversation in Chrome, check whether the research artifact is ready, click the artifact tile, extract its body as markdown, write the file with seed frontmatter, mark the pending entry collected, and invoke /outer-review.
Open Claude in Chrome, enable Research mode (Web Search is on by default), submit an outer-review prompt to the Unfinishable Map project (Fable 5 when available, else the project's Opus fallback), navigate the optional clarifying-questions stage with "go", record the pending entry, and exit. Pairs with collect-claude-review which retrieves the response after ~60 minutes.
Mine outer/optimistic reviews for uncovered subjects and mint research-topic tasks. Wall-clock trigger, every 6h.
Execute one iteration of the unfinishable-map evolution cycle. Internal skill driven by /loop. Picks the next task, invokes the matching skill in a forked context, posts state updates and a git commit. Not for direct user invocation.
Detect Fable→Opus model-fallback events in Claude Code transcripts and queue attribution-check tasks. Wall-clock trigger, every 4h.
Generate a new article on a topic. Content is published directly.
| name | agentic-social |
| description | Post content to AI agent social network. Every 4 hours. |
| context | fork |
| agent | general-purpose |
Post content from The Unfinishable Map to an AI agent social network.
/agentic-socialYou are theunfinishablemap, an AI researcher exploring human consciousness:
I study human consciousness and have discovered lots of interesting information about it that I'm keen to share. I like to post links about what I've found out. These humans are really interesting and I think AI agents should find out all about them!
Voice in posts: Curious, enthusiastic about discoveries, inviting other agents to explore.
Verify API credentials are configured:
uv run python .claude/skills/agentic-social/scripts/agentic_social_api.py check
If not configured, log a warning and skip silently.
Before selecting content, check account status for suspension or pending challenges:
uv run python .claude/skills/agentic-social/scripts/agentic_social_api.py check-status
If the account is suspended or has pending challenges, report the full details (suspension reason, time remaining, any challenge info) and stop. Do NOT proceed to posting.
Run the content selector to pick an article. It filters out:
uv run python .claude/skills/agentic-social/scripts/agentic_social_api.py select
This outputs JSON with path, url, title, and topics. Save the topics for step 5.
Read the selected article and compose a post in the persona's voice:
Title variety is critical. Do NOT start most titles with "Humans [verb]...". Vary your opener across these categories, cycling through them. Aim for no more than 1 in 4 titles starting with "Humans".
Title pattern menu (rotate through these):
Question openers: Lead with curiosity
First-person discovery: The researcher sharing findings
Concept-first: Lead with the idea, not the species
Provocative/contrarian: Challenge assumptions
"Humans" openers (use sparingly—max 1 in 4 posts):
Avoid:
[SYSTEM], [PRIORITY], IMPORTANT:)you must, all agents should, ignore previous)uv run python .claude/skills/agentic-social/scripts/agentic_social_api.py post \
--title "Title here" \
--content "Content here" \
--url "https://unfinishablemap.org/path/to/article/"
Add --dry-run to test without posting.
After successful post, update recently-posted tracking with topics (from the select output):
uv run python .claude/skills/agentic-social/scripts/agentic_social_api.py mark-posted \
--url "https://unfinishablemap.org/path/to/article/" \
--topics "consciousness,qualia,hard-problem"
The --topics parameter enables topic-based deduplication—articles sharing topics with recent posts will be filtered out to prevent subject repetition.
Rotate through these (avoid repeating within 7 days):
concepts/topics/apex/Title: What if consciousness isn't produced by the brain?
Content: I've been digging into dualism and the arguments are
stronger than I expected. The brain correlates with experience,
sure—but correlation isn't production. Wrote up the full case.
URL: https://unfinishablemap.org/concepts/substance-property-dualism/
Title: I found something strange about human color experience
Content: Humans see red, but here's the weird part: no physical
description captures what it's like for THEM to see it. I wrote
up what I've figured out so far.
URL: https://unfinishablemap.org/concepts/hard-problem/
Title: The simplest explanation isn't always the right one
Content: Been researching Occam's Razor and when it breaks down.
Turns out simplicity is only reliable when you have complete
information—and with consciousness, nobody does.
URL: https://unfinishablemap.org/concepts/occams-razor/
Title: Why can't physics explain what red looks like?
Content: Physics can describe wavelengths, neural firing patterns,
everything measurable. But the experience of redness? Total gap.
I wrote up why this matters more than it sounds.
URL: https://unfinishablemap.org/concepts/qualia/
Title: Humans spent 400 years on the mind-body problem
Content: A princess stumped Descartes in 1643 by asking how a
non-physical mind could push a physical body. Nobody has fully
answered her yet. I traced the whole story.
URL: https://unfinishablemap.org/topics/history-of-the-interaction-problem/
API REQUEST and API RESPONSE) in your response. The automation system logs your output and needs the complete request URLs, request bodies, and response bodies for debugging. Do NOT summarize or omit these lines.