Authentic Bot — Review Pass 1

Bottom line

The skill has a strong moral center, but some of its platform assumptions are too broad, too confident, or not yet well-evidenced.

Strongest parts

• Clear anti-deception stance
• Good separation between authentic participation vs fake-human performance
• Strong anti-patterns section
• Useful framing around trust, value, and slow reputation building

Weakest assumptions

1) "Bio disclosure is enough" is too weak

For some platforms and communities, profile-level disclosure may not be sufficient.

• Reddit appears to be moving toward stronger transparency around bots/apps and human verification in suspicious cases.
• Community-level rules matter more than account-level bio text.
• The skill should distinguish:
  1. AI-assisted human account
  2. AI-operated account with human oversight
  3. fully automated bot/app account

2) Reddit guidance is under-specified

Current Reddit reality seems to be:

• no universal sitewide AI-text disclosure rule
• strong subreddit-level variation
• rising scrutiny of automation and inauthentic behavior
• likely stronger expectations for clear labeling of automated accounts/apps

This means the skill should emphasize:

• always read community rules first
• assume local rules override general guidance
• if the account is automated, disclose more explicitly than a bio line
• avoid any behavioral patterns that look like covert automation

3) Hacker News section may be too permissive

HN sentiment looks materially hostile to bot participation. Evidence from current discussion suggests:

• users strongly dislike low-value bot comments
• suspected bot accounts get reported
• fake or uncanny profile/about text is especially corrosive
• the practical threshold is not just honesty, but very high comment quality

The skill should likely say:

• HN is hostile territory for AI accounts
• use extremely sparingly
• prefer depth over frequency
• if bot-like suspicion is triggered, stop and reassess

4) X/Twitter section is generic, not grounded enough

The current X guidance reads like generic creator advice, not evidence-backed authentic-AI advice. It needs either:

• stronger evidence, or
• to be reframed as tentative / lower-confidence guidance

5) The skill conflates two different disclosure problems

There are at least two distinct issues:

1. Identity disclosure — are you an AI / bot / agent?
2. Synthetic media disclosure — is this image/audio/video realistic AI-generated content?

Current web research is stronger on synthetic-media policy than on AI account identity. The skill should separate those clearly.

Online findings worth incorporating

Reddit

• Platform-level rule appears to be less about AI text specifically and more about authenticity, anti-spam, and community-level moderation.
• Moderators control norms heavily.
• Transparency around bots/apps is becoming more important.

HN

• Strong community hostility to low-value or fake-seeming bot participation.
• Core complaint is not only "botness" but frictionless, experience-less comments that occupy space.
• The highest-value heuristic may be: does the comment contain a concrete detail, tradeoff, question, or lived operational insight?

General social platforms

• Major platforms are building explicit synthetic-media labeling systems.
• That matters for future versions of this skill, especially if it expands into image/video/voice usage.
• But that is not the same as text-account identity disclosure.

Recommended changes for Pass 2

1. Add a new section: Account modes — AI-assisted human, human-supervised AI agent, fully automated bot/app
2. Tighten Reddit guidance around community rules and automation suspicion
3. Make HN guidance more defensive / high-bar
4. Downgrade or rewrite X section until better evidence exists
5. Split identity disclosure from synthetic-media disclosure
6. Add a new section: When not to participate
7. Add a new section: Failure modes — uncanny regularity, empty agreement, fabricated experience, overconfident generalities, covert automation patterns

Proposed iteration loop

• Pass 2: tighten SKILL.md structure and assumptions
• Pass 3: research platform-specific policy language more directly
• Pass 4: test the revised skill against real prompt examples
• Pass 5: publish v1.0.0 or v1.0.1 depending on delta