Loading ToolBench
Channel-safety layer for cold outreach on email, LinkedIn, Instagram DM, X DM: volume limits, SPF/DKIM/DMARC sender setup, spam-trigger auditing, benchmark interpretation, campaign failure diagnosis.
Server provides 6 well-intentioned tools for outreach safety with clear domain focus. All tools have descriptions (194 - 350 chars, within baseline 34 - 392 range) and input schemas with type declarations. However, output schemas are entirely undocumented, LLMs cannot predict response structure or chain tools. Parameter descriptions are present but generic; many lack constraints (enums, ranges, formats). Tool names are descriptive but inconsistent in verb usage (channel_rules, sender_setup_checklist lack action verbs). No error handling guidance, no recovery paths, no validation rules documented. Security posture is read-only (low risk), but descriptions do not state this explicitly for each tool.
Audits email subject and body text for spam phrases, structural red flags (ALL-CAPS, excessive punctuation, multiple links, images, money symbols, fake Re:/Fwd: prefixes), and returns a spam score (0-100) with hit details and remediation advice.
Interprets email campaign metrics (open rate, reply rate, bounce rate, spam complaint rate) against channel benchmarks and returns interpretation, risk assessment, and next-step recommendations.
Returns safe volume limits, account setup, warmup, kill signals, personalization floor, and legal basics for email, LinkedIn, Instagram DM, or X DM.
Diagnoses why a campaign is failing by analyzing symptoms (low open rate, high bounce, account restrictions, etc.) and returns root-cause hypotheses, evidence checklist, and recovery steps.
Returns a 10-step email infrastructure checklist: lookalike domains, SPF, DKIM, DMARC, mailbox provisioning, warmup, list verification, tracking domain, unsubscribe/address, and monitoring.
Returns a safe volume ramp schedule for a new mailbox or account on a given channel, accounting for warmup, list quality, and platform throttling. Includes daily/weekly targets and kill-switch thresholds.
No output schemas documented. LLMs cannot predict response structure, field names, or types. Breaks tool chaining and forces agents to guess at downstream parameter mappings.
Tool names lack consistent action verbs. 'channel_rules' and 'sender_setup_checklist' are nouns; should be 'get_channel_rules' and 'get_sender_setup_checklist'. LLMs infer intent from verb-noun structure.
Parameter 'channel' in multiple tools accepts free-form string with enum constraint only in description ('One of: email, linkedin, instagram_dm, x_dm'). Should be declared as enum type in schema, not prose.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | F | 44 | 2026-07-28+ | v2 |
No error handling guidance. Tools return read-only data, but descriptions do not state this explicitly. No recovery paths documented if invalid channel or malformed input provided.
Parameter descriptions lack validation rules. E.g., 'open_rate' in benchmark_interpretation accepts 0-100 but no min/max constraint stated. 'account_age_days' in volume_ramp_plan has no range guidance.