Development record

The grill hands the interview to an installed interviewing skill, and runs its own when there is none

intentdraftsminoritd-2609292108373494

Typed links: related_issues iss-165 (the record this draft plans); related_adrs adr-22 (native default, an external tool as an opt-in adapter).

Press Release

When the person has installed an interviewing skill they prefer, abcd's grill hands the interview to it; when they have not, abcd runs its own full interview. The hand-over is visible and opt-in: abcd says which interviewer is asking, the installed skill is never required, and the record abcd keeps from the interview has the same shape whichever interviewer asked. "I already use an interviewing skill I trust everywhere else," said Iris, a product thinker. "Now abcd uses it too, and on a machine without it I still get abcd's own interview rather than none."

Why This Matters

abcd's interviews (the grill of a draft and the planning interview) run on abcd's own question rules, the GRILL rule domain. A person who already keeps an interviewing skill they trust in their host gets a second, different interviewer inside abcd. iss-165 asks for the pattern abcd uses elsewhere, basics built in and the state of the art delegated: hand the interview to the installed skill when it is present, and run abcd's own comprehensive interview when it is not. The product thinker ruled on 2026-09-29 to plan it (ruling J11).

What's In Scope

Detection of an installed interviewing skill the person has named, at the moment an interview starts.
The hand-over: the interview runs through that skill, and abcd says which interviewer is asking.
The fallback: with no such skill, abcd's own interview runs unchanged.
One record shape: what abcd keeps from the interview (the draft's edited sections, the answers it records) is the same whichever interviewer asked.

What's Out of Scope

Shipping or vendoring any third-party skill: the installed skill is an opt-in dependency, never a requirement.
The sign-off act itself: abcd intent plan stays the product thinker's act, whichever interviewer ran the questions.

Mechanism

Prompted (the claim-recording gradient): why the authors expect this to work, as a falsifiable "we expect X because Y" — not the outcome restated. Replace this line with the claim, or with the exact token None stated. alone on its line to record the claim as considered and declined.

Scope Conditions

Required (the claim-recording gradient): the population, platform, scale, or assumptions this claim holds under, one per top-level bullet — abcd intent plan stamps each with a persistent identity. Replace this line with those bullets, or with the exact token None stated. alone on its line.

Acceptance Criteria

Seeded by the run from iss-165 and the ruling, unconfirmed: the planning interview walks each one.

Given a host with the named interviewing skill installed, when an interview starts, then abcd hands the interview to that skill and says so before the first question.
Given a host without it, when an interview starts, then abcd's own interview runs and says which interviewer is asking.
Given an interview run by either interviewer, when it ends, then the draft's recorded sections and the planning gate's inputs have the same shape.
Given a hand-over that fails part-way, when the interview resumes, then abcd's own interview continues from the last recorded answer and says that it took over.

SOTA

Surveyed on 2026-10-23; sources, evidence tiers and the local fit risks are in the elicitation interviews SOTA note.

A skill has a small, stable on-disk shape: SKILL.md with name (which "Must match the parent directory name") and description, optional compatibility, and progressive disclosure (Agent Skills specification). Detection reads those two fields and one more: a skill with disable-model-invocation: true cannot be invoked by the agent on its own (its description is not even in context), so such a skill is detected but cannot be handed over, and abcd tells the person to type the skill's command or runs its own interview.
Where a host looks differs per host: the host abcd ships for reads personal, project, nested, additional-directory and plugin skill directories, and other hosts use their own (one 2026 survey lists six). The path table is the part a new host replaces.
abcd can start a skill only through the host (its CLI or SDK loads a leading /skill-name), and a headless start cannot hold an interview: there the host's question tool auto-resolves with an empty answer (anthropics/claude-code issue 50728, closed "not planned"), and even interactively it has an open rendering bug (February 2026). A hand-over is therefore only meaningful inside a live interactive session, started by abcd's command page telling the agent to invoke the named skill.
Public interviewing skills disagree on abcd's core rule: the brainstorming skill forbids option menus ("Options steer"), a grill-me skill gives "your recommended answer" with each question. abcd's rules travel to the external interviewer as a stated brief, abcd records which interviewer asked, and the record shape is unaffected because abcd writes it from the answers.
The skill ecosystem is unvetted (a vendor scan put critical issues in 13.4 per cent of 3,984 skills; the vendor sells the remedy): detect one skill the person named, never scan for "any interviewing skill".

Path 2: the person names one skill (user-level, overridable per repository); abcd checks the host-specific path for a SKILL.md whose name matches its directory and that does not set disable-model-invocation; the command page hands over in a live session and prints the asking discipline it expects; otherwise, and always when headless, abcd's own interview runs and says so. Covers the grill and the planning interview alike. No Go dependency.

The independent fit-challenge runs at plan time.

Open Questions

How the person names the skill: a setting in the repository's configuration, a user-level setting, or detection by name.
Which of abcd's question rules bind the external interviewer: one question at a time and the options that widen are abcd's stance; does the hand-over pass them to the skill, or accept the skill's own?
Which interviews it covers: the grill of a draft only, or the planning interview too.
(SOTA 2026-10-23) Whether answers taken by an interviewer that recommends (grill-me's shape) need a marker in the record, given widen-options-never-recommend binds abcd's own questions only.

Audit Notes

Empty. Populated by intent-auditor when intent moves to shipped/.

Frontmatter

id	itd-2609292108373494
type	intent
lifecycle	drafts
severity	minor

.abcd/development/intents/drafts/itd-2609292108373494-when-the-person-has-installed-an.md
open on GitHub ↗ · commit history ↗

References1

adr-22Bundled dependencies become pluggable adapters over a native defaultrelated

Referenced by4

iss-165The grill/interview capability should follow the basics-built-in, SOTA-delegated pattern: when an external state-of-the-art interviewing skill is installed in the host (the maintainer names Matt Pocock's grill skill), abcd's grill delegates to it; when absent, abcd offers its own comprehensive built-in interview instead. Same stance as the oracle boundary: the external skill is an opt-in dependency, never a requirement.related
iss-2610262034051312draft itd-2609292108373494 hands over 'abcd's grill'; the intent surface says no grill sub-verb shipsrelated
iss-2610262040422553itd-2609292108373494 calls the planning interview 'abcd's grill'; the interview glossary says the interviews that ship are not grill sessions.related
itd-2610030810350727Every question abcd asks reads the same way, on a narrow window and a wide onerelated

mentions3

dcn-26092900000000112026-09-29 — The product thinker answered the twenty-six Group J rulings (plan or close), J1 to J26 (recorded by lane recGroupJ of autonomous run A, for orchestrator abcd-fc). The answers are kept verbatim in the local-tier file `.abcd/.work.local/scratch/reports/rulings-answered-2609-29.md`, section "Group J", dated 2026-09-29, and are restated here because that file is not committed. PLAN: the installer script becomes a minimal starter with an ownership-checked hand-over, planned as a new intent with its planning interview owed (J2, iss-377, draft itd-2609292106557115); token metering and size classes are planned as one intent (J3, iss-2608301744251874 and iss-2608301856299268), and session token accounting in the history store joins that metering intent (J4, iss-2608220150157508; all three in draft itd-2609292107351737); the sources tooling moves into abcd's core with ingest and consult as abcd verbs (J5, iss-27, whose corpus half itd-76 already delivered), and the backfill of the placeholder source entries is part of that intent (J6, iss-55; both in draft itd-2609292108089653); the grill hands off to an installed external interview skill and falls back to its own (J11, iss-165, draft itd-2609292108373494); an opt-in local-model prompt sanitiser is planned (J14, iss-2608261543489261, draft itd-2609292109005937); decision-to-transcript links are planned, with the anchor kind and the link's home settled in the planning interview owed for draft itd-171 (J16, iss-2608290819228175); the dredge synthesis, its write-ups as their own memory source class, is planned together with draft itd-25, whose planning interview is owed (J19, iss-2609211905340006); one provenance register across ingest and vendoring is planned together with draft itd-26, whose planning interview is owed (J20, iss-2609211905346507); release retention, keeping the newest release per version line, is automated by planning draft itd-70 (J21, iss-282); a rules-backend seam with an opt-in CARL adapter is planned, the native loader staying the default (J22, iss-64, draft itd-2609292109214516); core-owned managed pre-commit gates are planned with draft itd-62 (J23, iss-84); a behavioural end-to-end scenario suite per verb family is planned (J24, iss-48, draft itd-2609292109475690). Every draft stays in drafts/ until a person plans it, and each source issue links its draft and names that planning interview in its deferral. BUILD: the teaching plane, the rules-loader safety domain generated from the hazard registry, is built as a lane, the registry being the single source (J10, iss-151 and itd-103; a code lane owns it). RULED DETAILS: `source_kind` carries both, as two separate labels, one for the tool and one for the route (J13, iss-2608230752354928); the prompt router's removal signal is the full list of active domains every time, a domain's absence meaning it has stopped (J15, iss-2608261550580260); setup offers an outside AI service (for example OpenRouter) at install, skippable, once the API adapter ships (J18, iss-2609081951416843, which waits on the API adapter); the salvage-hook timeout is 120 seconds, the wait must show a counter or a message while it runs and never stall silently, and every per-event timeout is pinned in a test (J25, iss-323). J13, J15 and J25 are carried by code lanes. CLOSE: the quick-tunnel preview protocol is closed unless the need recurs (J26, iss-2608230617385431, wontfix with that reason). PARKED ON A TRIGGER: agent payment protocols are kept under watch, not now, revisited as the protocols mature (J8, iss-138); the findings-only skill waits for the skill format to be versioned and stewarded (J9, iss-139); the disposition worksheet as an intent and a site page waits until after the first study (J17, iss-2609021815529020). DECIDE LATER: the Homebrew tap stays parked (J1, iss-380); pluggable search back ends (J7, iss-26); forge-backed record numbering (J12, iss-2608210737264758).
dcn-26092900000000202026-09-29 — The product thinker answered the twenty-six Group J rulings (plan or close), J1 to J26 (recorded by lane recGroupJ of autonomous run A, for orchestrator abcd-fc). The answers are kept verbatim in the local-tier file `.abcd/.work.local/scratch/reports/rulings-answered-2609-29.md`, section "Group J", dated 2026-09-29, and are restated here because that file is not committed. PLAN: the installer script becomes a minimal starter with an ownership-checked hand-over, planned as a new intent with its planning interview owed (J2, iss-377, draft itd-2609292106557115); token metering and size classes are planned as one intent (J3, iss-2608301744251874 and iss-2608301856299268), and session token accounting in the history store joins that metering intent (J4, iss-2608220150157508; all three in draft itd-2609292107351737); the sources tooling moves into abcd's core with ingest and consult as abcd verbs (J5, iss-27, whose corpus half itd-76 already delivered), and the backfill of the placeholder source entries is part of that intent (J6, iss-55; both in draft itd-2609292108089653); the grill hands off to an installed external interview skill and falls back to its own (J11, iss-165, draft itd-2609292108373494); an opt-in local-model prompt sanitiser is planned (J14, iss-2608261543489261, draft itd-2609292109005937); decision-to-transcript links are planned, with the anchor kind and the link's home settled in the planning interview owed for draft itd-171 (J16, iss-2608290819228175); the dredge synthesis, its write-ups as their own memory source class, is planned together with draft itd-25, whose planning interview is owed (J19, iss-2609211905340006); one provenance register across ingest and vendoring is planned together with draft itd-26, whose planning interview is owed (J20, iss-2609211905346507); release retention, keeping the newest release per version line, is automated by planning draft itd-70 (J21, iss-282); a rules-backend seam with an opt-in CARL adapter is planned, the native loader staying the default (J22, iss-64, draft itd-2609292109214516); core-owned managed pre-commit gates are planned with draft itd-62 (J23, iss-84); a behavioural end-to-end scenario suite per verb family is planned (J24, iss-48, draft itd-2609292109475690). Every draft stays in drafts/ until a person plans it, and each source issue links its draft and names that planning interview in its deferral. BUILD: the teaching plane, the rules-loader safety domain generated from the hazard registry, is built as a lane, the registry being the single source (J10, iss-151 and itd-103; a code lane owns it). RULED DETAILS: `source_kind` carries both, as two separate labels, one for the tool and one for the route (J13, iss-2608230752354928); the prompt router's removal signal is the full list of active domains every time, a domain's absence meaning it has stopped (J15, iss-2608261550580260); setup offers an outside AI service (for example OpenRouter) at install, skippable, once the API adapter ships (J18, iss-2609081951416843, which waits on the API adapter); the salvage-hook timeout is 120 seconds, the wait must show a counter or a message while it runs and never stall silently, and every per-event timeout is pinned in a test (J25, iss-323). J13, J15 and J25 are carried by code lanes. CLOSE: the quick-tunnel preview protocol is closed unless the need recurs (J26, iss-2608230617385431, wontfix with that reason). PARKED ON A TRIGGER: agent payment protocols are kept under watch, not now, revisited as the protocols mature (J8, iss-138); the findings-only skill waits for the skill format to be versioned and stewarded (J9, iss-139); the disposition worksheet as an intent and a site page waits until after the first study (J17, iss-2609021815529020). DECIDE LATER: the Homebrew tap stays parked (J1, iss-380); pluggable search back ends (J7, iss-26); forge-backed record numbering (J12, iss-2608210737264758).
itd-2610230756259215Every question abcd asks passes its check before the person sees it