Anti-hype Filter

Detect hype cycles and neutralize emotional triggers by rewriting claims into verifiable structures and explicit risk/uncertainty.

Security Scan

The skill is instruction-only and its declared inputs, outputs, and steps match the stated goal of detecting and rewriting hypey language; it requests no credentials, makes no external calls, and has no install footprint.

Purpose & Capability

Name and description match the SKILL.md steps (detect triggers, classify, rewrite). There are no unexpected environment variables, binaries, or config paths requested that would be unrelated to a text-filtering/rewriting tool.

Instruction Scope

Runtime instructions are narrowly scoped to processing provided text (extract claims, detect triggers, classify, rewrite, draft response). The SKILL.md does not instruct reading system files, environment variables, or sending data to external endpoints, and includes sensible safety rules (avoid accusing individuals, no fabricated data).

Install Mechanism

No install spec or code files requiring downloads or execution are present—this is an instruction-only skill, so nothing is written to disk or installed at runtime.

Credentials

The skill requires no environment variables, credentials, or config paths. The declared inputs (text, optional policy/hype_terms) are proportional to the stated purpose.

Persistence & Privilege

always is false and the skill does not request elevated persistence or modification of other skills. Autonomous invocation is allowed by default but that is expected for normal skills and is not combined with other red flags here.

Assessment

This skill appears internally consistent and low-risk, but before installing: (1) avoid feeding sensitive or confidential text (it rewrites whatever you send); (2) review or supply the policy.hype_terms list to ensure it matches your domain and avoids false positives/negatives; (3) test on representative examples to confirm rewrites don't inadvertently change factual meaning or introduce bias; (4) consider logging/auditing outputs if you will apply this automatically to user-facing content; and (5) if you plan to enable autonomous invocation, restrict its scope and monitor outputs to prevent undesired automated moderation decisions.

SKILL: anti-hype-filter

Purpose

Detect and neutralize hype cycles before they distort system integrity by stripping emotional triggers and replacing them with structural analysis.

When to Use

• "guaranteed", "moon", "100x", "alpha" style language
• Urgency without substance ("now or never")
• Social proof without evidence
• Claims that minimize risk or constraints

Inputs

• text (required): message to evaluate
• context (optional):
  • domain (token|product|governance|community)
• policy (required):
  • hype_terms (optional list; if omitted, use the embedded default set in this skill)
  • max_response_words (default 100)

Steps

1. Extract key claims (1-5).
2. Detect hype triggers:
   • urgency framing
   • certainty language
   • vague upside claims
   • social proof substitution
3. Classify:
   • signal, noise, or manipulation_risk
4. Rewrite the message into a verifiable form:
   • replace certainty with uncertainty
   • add required missing variables (data window, metrics, constraints)
5. Draft a minimal response that:
   • does not repeat hype memes verbatim
   • demands evidence and risk disclosure

Validation

• If classification is manipulation_risk, provide at least 1 falsifiable request for evidence.
• Do not amplify hype phrases; paraphrase instead.

Output

• anti_hype_result:
  • classification ("signal"|"noise"|"manipulation_risk")
  • detected_triggers (list)
  • missing_information (list)
  • rewrite (verifiable version)
  • response_draft (string)

Safety Rules

• Never accuse individuals of malice without evidence; label as "risk" not "intent".
• No financial promises.
• No deception; no fabricated data.

Example

Input: "This will 100x in 2 weeks, everyone knows." Output: manipulation_risk, missing evidence, rewrite into metrics/timeframe/assumptions, and a short demand for proof + risk disclosure.