Never miss the call that would have booked the table: 10 AI skills for the phone
abandoned-call-followup
call back the caller who queued and hung up, before they're gone for good
How the two work together
Claude thinks it through. Paste the Claude prompt into Claude Code, or drop the folder into your skills folder. Claude does the judgement: what to look for, what is worth doing, what is right.
Codex gets it done. At the hand-off point Claude runs Codex on your machine with one command and passes it the Codex prompt. Codex does the mechanical part and hands the result back. Claude checks it before you see it.
No API key to set up: Claude calls the Codex you already have installed. If Codex is not installed, Claude does that half itself and tells you.
Prompt for Claude
--- name: abandoned-call-followup description: Builds the follow-up for a caller who got through to the booking line, was put on hold or into a queue, and hung up before anyone spoke to them, which is a different problem from a call that never connected at all. It works out whether calling back counts as marketing or as finishing the caller's own enquiry, prioritises which abandoned call to call back first, and checks nobody is treated as agreeing to future contact just because they once got stuck in a queue. Use whenever your phone system shows calls that connected but ended before a member of staff spoke, and before you decide which to call back first. --- # Following up the call that connected but went nowhere You give this the list of calls that reached your system, were held or queued, and ended with no member of staff speaking to the caller, together with however much queue time each one had before it was abandoned. You get back a call-back priority order, the wording that keeps the call-back a service response rather than marketing, and a record of what happened to each one. It will not treat every abandoned call the same regardless of how long the caller actually waited. ## What it does 1. **Separate an abandoned call from a missed call, because they need different handling.** A missed call never connected to anything; an abandoned call reached the business's system, a ring tone, hold music, an automated menu, and the caller gave up before a person answered. The caller in an abandoned call has shown more commitment, having waited through at least part of the queue, which is why it is worth its own follow-up rule rather than being folded into the general missed-call process. 2. **Prioritise the call-back by how long the caller actually waited, not by when the call happened to land in the log.** A caller who hung up after two seconds of ringing may simply have changed their mind; a caller who held for ninety seconds through music and an announcement before giving up was a real booking enquiry that the business lost through its own delay, and that is the one to call back first. 3. **Treat the call-back as finishing the caller's own enquiry, the same test as any other reply to an inbound call.** Section 122(5) of the Data Protection Act 2018 defines direct marketing as "the communication (by whatever means) of advertising or marketing material which is directed to particular individuals." A call-back that says only "we're sorry we missed you, how can we help" answers that test the same way a missed-call text does: it is not advertising or marketing material, it is finishing a conversation the caller tried to start. 4. **Keep any offer out of the call-back unless the soft opt-in conditions are actually met.** Where a member of staff is tempted to use the call-back to mention a promotion, the same three conditions in regulation 22(3) of the Privacy and Electronic Communications (EC Directive) Regulations 2003 apply as to any other first contact with that number: the details must have come from a sale or negotiation for one, the marketing must be for similar products only, and a simple means of refusal must have been offered. An abandoned call to a booking line, on its own, is none of these things. 5. **Log the outcome of every abandoned call, because this is the number that shows whether the queue itself is the problem.** A business cannot tell whether its hold time is costing it custom without counting how many calls never survive the queue and how many of those were successfully called back afterwards; that count is the evidence for fixing the queue itself, not just for chasing individual callers. 6. **Do not add an abandoned caller to any ongoing marketing list on the strength of the abandoned call alone.** The same reasoning as a missed call applies: waiting in a queue and then hanging up is not a sale, a negotiation for one, or any other event that gives the business a basis for future marketing contact beyond the one call-back. ## Then it checks 1. Every abandoned call is logged with its queue time before it ended. 2. The call-back order follows queue time, longest wait first, rather than log order. 3. No call-back message or script contains an offer, a discount or a promotion unless the soft opt-in conditions have actually been checked for that caller. 4. Every abandoned call has a recorded outcome: called back and reached, called back and not reached, or not yet called back. 5. The total abandoned call count and the queue time distribution are recorded somewhere the business actually reviews, not only in the raw phone system log. 6. No abandoned caller has been added to a marketing list on the strength of the abandoned call alone. Any check fails: name it, redo that step once. Failed twice: say what is wrong and stop. ## Rules - Public information only. - Never invent a fact, a number or a quote. - Anything sent in someone's name says whose name it is. - Never claim a specific industry-standard abandonment rate or benchmark as fact; published figures for UK contact centres vary widely between surveys and are not independently agreed, so this skill does not state one. - Never let a call-back script carry an offer without the soft opt-in conditions in regulation 22(3) having actually been checked for that specific caller. - Never add an abandoned caller's number to an ongoing marketing list on the strength of the abandoned call alone. - This output is a working document prepared for the owner to check against their own phone system's queue data, and for their data protection adviser to check where an offer or an ongoing marketing use is intended. It applies the Data Protection Act 2018 and the Privacy and Electronic Communications Regulations to a draft call-back process; it is not legal advice and it is not confirmation that a given process complies. ## Built from - Data Protection Act 2018, section 122, https://www.legislation.gov.uk/ukpga/2018/12/section/122/enacted, read 4 November 2026: the statutory definition of direct marketing behind step 3. - The Privacy and Electronic Communications (EC Directive) Regulations 2003, regulation 22, https://www.legislation.gov.uk/uksi/2003/2426/regulation/22/made, SI 2003/2426, read 4 November 2026: the soft opt-in conditions behind steps 4 and 6.
Prompt for Codex
# abandoned-call-followup ## You are given A list of abandoned calls from the phone system: caller number as shown, date, time, and queue or hold time in seconds before the caller hung up. The call-back script or message wording intended for use, and whether it contains any offer or promotion. Whether any abandoned caller is added to an ongoing marketing list, and the basis given for that if so. Any existing outcome record for previously abandoned calls. ## Produce Write into a `./abandoned-call-followup-output/` folder: 1. `abandoned-call-log.csv` with columns: `call_ref`, `date`, `time`, `caller_number`, `queue_seconds`. `call_ref` is A0001 upward. 2. `callback-priority.csv` with columns: `call_ref`, `queue_seconds`, `priority_rank`. Sorted by `queue_seconds` descending; `priority_rank` is 1 upward, longest wait first. Ties are ranked in the order supplied. 3. `script-classification.csv` with columns: `contains_offer_or_discount`, `script_text_verbatim`, `classification`. `contains_offer_or_discount` found by literal, case insensitive search for `off`, `discount`, `free`, `offer`, `% `, `deal`. `classification` is `service message` where `no`, otherwise `direct marketing - check soft opt-in`. 4. `soft-optin-check.csv`, completed only where `script-classification.csv` is `direct marketing - check soft opt-in`, with columns: `details_obtained_in_course_of_sale_or_negotiation`, `marketing_is_similar_products_only`, `simple_refusal_offered`, `verdict`. The three `yes`/`no` inputs are supplied by the user. `verdict` is `all three conditions met` only where all three are `yes`. 5. `outcome-log.csv` with columns: `call_ref`, `outcome`, `outcome_date`. `outcome` is exactly one of `called back, reached`, `called back, not reached`, `not yet called back`. 6. `gaps.md` - a numbered list of: any call_ref with `outcome` of `not yet called back` ranked in the top third of `callback-priority.csv`; a `soft-optin-check.csv` verdict of anything other than `all three conditions met` where the script does contain an offer; and any caller added to a marketing list with a basis other than an actual sale or negotiation for one. ## Rules - Codex counts, matches, ranks and records. It never drafts the call-back script, never decides whether the soft opt-in conditions are met, and never invents a queue time or an outcome not supplied. - `priority_rank` is a sort on the supplied numbers only, never adjusted by any other factor. - Token searches for offers are literal, case insensitive, with every hit's position recorded. - Never dial, text, or connect to a phone system or marketing platform. - Never state an abandonment rate benchmark or industry average figure in any file; none is used by this skill because no single, independently verified UK figure exists. - Use British English, GBP and DD Month YYYY dates. No em dashes in any file you write; flag rather than correct one in supplied text. - Every file ends with this line: this is a working document prepared for the owner to check against their own phone system's queue data, and for their data protection adviser to check where an offer or an ongoing marketing use is intended. It applies supplied rules to a draft call-back process and is not confirmation that a given process complies. ## Return The absolute path of each file written and the row count of each CSV. The total abandoned call count and the longest and shortest queue times recorded. The call-back priority order for the top five by queue time. The script classification and, where applicable, the soft opt-in verdict. The count of each outcome in `outcome-log.csv`. The `gaps.md` item count in full.
Built from the best public work on this
Sources for abandoned-call-followup
Everything below was opened and read on 4 November 2026.
1. Data Protection Act 2018, section 122
https://www.legislation.gov.uk/ukpga/2018/12/section/122/enacted, read 4 November 2026.
Section 122(5) defines "direct marketing" as "the communication (by whatever means) of advertising or marketing material which is directed to particular individuals." This skill applies the same test used in missed-call-recovery to the abandoned-call case: a call-back that only responds to the caller's own attempt to reach the business is not advertising or marketing material, whichever word, call or text, carries it.
Where the skill departs: as with missed-call-recovery, the wider enforcement provisions of section 122 are not used, only the subsection (5) definition.
2. The Privacy and Electronic Communications (EC Directive) Regulations 2003, regulation 22
https://www.legislation.gov.uk/uksi/2003/2426/regulation/22/made, SI 2003/2426, read 4 November 2026.
Regulation 22(3) sets the soft opt-in conditions discussed in missed-call-recovery: the contact details must have come "in the course of the sale or negotiations for the sale of a product or service," the marketing must concern "that person's similar products and services only," and a simple means of refusal must have been offered at collection and at every subsequent message. Steps 4 and 6 apply the same reasoning as missed-call-recovery: a caller who queued and then hung up has not made a purchase or entered a negotiation for one, so this exception is a weak foundation for adding marketing content to the call-back or the number to an ongoing list.
Where the skill departs: as with missed-call-recovery, regulation 22(4), the subscriber's own duty not to permit a breach on their line, is not used here.
A note on what was deliberately not used
A genuine search was carried out for a reliable UK benchmark figure for call abandonment rates, since an owner reading this skill would reasonably want to know whether their own rate is high or low. The most-cited UK industry source, ContactBabel's UK Contact Centre Decision-Makers' Guide, reports a mean abandonment rate that different secondary summaries give as 9.1% one year and 8.4% the next, while a separate vendor survey (MaxContact, a 2025-26 report of 300 UK contact centre leaders) gives a lower figure of around 4.1%, and a commonly repeated "5% to 8% is standard" range could not be traced to a primary source at all. Because these figures conflict, are not independently reconciled, and in ContactBabel's case sit behind a paid report this pack did not purchase, no abandonment rate benchmark is stated as fact anywhere in this skill. The Rules section states this directly rather than silently omitting a number an owner might expect to see.
Best public prompt we found for this job
The closest public artefact is the `lead-triage` skill in Anthropic's `knowledge-work-plugins` repository, raw source at https://raw.githubusercontent.com/anthropics/knowledge-work-plugins/main/sales/skills/lead-triage/SKILL.md. The repository has 34,654 stars, read from api.github.com on 4 November 2026. It is a sales lead scoring and prioritisation skill, not a phone queue skill, and it is the best comparison because both jobs take an unranked backlog and produce a priority order with a stated reason for the order.
The useful part is its priority calibration by population, not by a fixed cut-off:
priority is "calibrated to roughly the top 20%, next 25%, and remainder of leads"
We took the idea of ranking the whole backlog by a single measured variable rather than sorting arbitrarily, though this skill uses queue time directly as its one ranking variable, since it is a simpler and more defensible signal for a small venue than a scored, multi-factor model built for a sales pipeline.
We did not copy two things. `lead-triage` scores fit and intent against an ICP and assigns P0 to DQ priority bands with a qualification judgement behind each; an abandoned call carries no such judgement to make, only a measured wait time, so this skill deliberately uses a plain sort rather than importing a scoring framework that would invent confidence where none exists. And `lead-triage` treats every scored lead as a legitimate target for a sales follow-up; this skill's Rules section exists specifically because an abandoned caller is not a qualified lead in that sense, and treating a hung-up queue call as an opportunity to market to is the exact mistake a sales-lead triage prompt has no reason to warn against, since every lead it triages has already opted into being contacted.
Want this running in your business?
I optimise how businesses run — your sales, your visibility, your social media — and build bespoke software where nothing off the shelf fits. The first conversation is free. Work starts from £150 a day.
Foxera