| name | prompt-grill-me |
| description | Log a corrective interaction as plugin feedback, then grill the fix inline — a bounded one-question-at-a-time interrogation (≤5 questions) of the correction following the embedded grilling technique. Self-contained; no plugin dependency. Activated when the user prompt starts with "prompt-grill-me:".
|
| allowed-tools | view, edit, create, bash, glob, grep, task, ask_user |
Log a corrective interaction, then grill the fix: the argument (text following the prompt-grill-me: trigger)
prompt-grill-me: is for when a dev-workflows command produced something wrong
and you want a bounded one-question-at-a-time interrogation (≤5 questions) of
the correction. It captures the corrective triple as plugin feedback, then
grills the fix inline following the embedded grilling technique
(~/.copilot/installed-plugins/ihudak-copilot-plugins/dev-workflows/skills/_shared/grilling-technique.md). origin: prompt.
The interrogation is self-contained — this command owns the grill and has no
plugin dependency.
Phase 1 — Identify the target
Infer the target command from recent context — which command's output you are
correcting. Ask only if genuinely ambiguous (one grouped prompt, last choice
"Other… (describe)"). If no command applies, use n/a.
Phase 2 — Persist the corrective triple
Cite ~/.copilot/installed-plugins/ihudak-copilot-plugins/dev-workflows/skills/_shared/feedback-emission.md and call its
emit-prompt entry point (§6). Provide:
- Friction — what the command produced that was wrong.
- User prompt — the
prompt-grill-me: argument, verbatim (never paraphrased).
- Resolution —
Grilled the fix inline.
command (Phase 1), an inferred category (§1 vocab, reuse-first), impact,
jira_key (or null), source.
emit-prompt resolves the write target via the §2 specs-first ladder, formats
the entry with the two extra prose blocks (origin: prompt), appends per §3
(never silently skipped), and writes silently. Surface the persisted path.
Phase 3 — Grill the fix (inline)
Interrogate the correction directly, following
~/.copilot/installed-plugins/ihudak-copilot-plugins/dev-workflows/skills/_shared/grilling-technique.md:
- Depth: bounded — a capped set (≤5) of the highest Impact×Uncertainty
questions about the fix, then stop; record any leftover high-impact gaps.
- Stage: the correction itself — why the original output was wrong, what the
right shape is, and what should change so the mistake does not recur.
Follow the technique's mechanics (one question at a time, a recommended answer
each time, fact-vs-decision split, dependency order). This command NEVER
commits, and NEVER writes into a docs/code repo or the current working
directory.
