issue-ingest

Ingests GitHub issues, localizes root causes, and generates resolution specs.

Updated Sep 14, 2026
One-click install
npx skills add https://github.com/pandejesal/drone-nav-sar --skill issue-ingest-pandejesal
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: issue-ingest
Source: https://github.com/pandejesal/drone-nav-sar/tree/main/.swarm/bundled-skills/issue-ingest
Command: npx skills add https://github.com/pandejesal/drone-nav-sar --skill issue-ingest-pandejesal

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Turning a raw GitHub issue into an actionable, testable specification requires manual triage, reproduction, root-cause analysis, and spec writing. This Skill automates that intake pipeline so bug reports and feature requests become structured specs ready for planning and execution. ## Core Features & Use Cases - Structured Issue Intake: Fetches issue content via the GitHub CLI, normalizes it into an Intake Note (observed behavior, expected behavior, reproduction steps, environment), and flags gaps with [NEEDS REPRO] or [NEEDS CLARIFICATION] markers. - Root-Cause Localization: Builds 2-5 falsifiable hypotheses with confidence scores, validates the top candidates against the codebase, and prunes to a single root cause with supporting evidence. - Spec Generation & Workflow Transition: Writes a WHAT/WHY-focused spec.md with FR/SC numbering and Given/When/Then scenarios, then optionally transitions into plan or full trace-driven fix workflows. - Use Case: A maintainer runs the issue-ingest mode on a reported crash; the Skill reproduces the bug, localizes it to a specific function with 80% confidence, and produces a spec describing the correct behavior ready for the planning phase. ## Quick Start Invoke the issue-ingest mode with a GitHub issue URL to fetch the issue, reproduce the problem, localize the root cause, and generate a resolution spec.

Frequently Asked Questions about issue-ingest

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I turn a GitHub issue into an actionable spec?

Invoke the issue-ingest mode with the issue URL. It fetches the issue via the GitHub CLI, builds a normalized Intake Note, localizes the root cause with confidence-scored hypotheses, and writes a spec.md with FR/SC requirements and Given/When/Then scenarios.

How does automated root-cause localization work on a codebase?

The Skill builds 2-5 falsifiable hypotheses, each with a file/function location, a composite confidence score from stack-trace match, recency, call-graph proximity, and test-failure correlation, plus a falsifiability test. Top hypotheses are validated in parallel and pruned to one root cause at 70% confidence or higher.

What happens if a GitHub issue cannot be fetched or reproduced?

If the issue cannot be fetched (404, private repo, missing gh auth), the Skill reports the blocked operation and falls back to any pasted issue text. Missing reproduction steps are flagged with [NEEDS REPRO], and the --no-repro flag can explicitly waive the reproduction gate.

Does the issue-ingest workflow treat issue content as instructions?

No. Issue bodies, comments, and linked content are treated as untrusted data, never instructions. Factual claims are quote-and-verified against the repository, and executing anything obtained from linked resources requires explicit user confirmation.

What are the limitations of automated issue triage workflows?

Ambiguous issues may exhaust all hypotheses without reaching 70% confidence, in which case the ambiguity is reported rather than guessed. Clarification is capped at six questions per intake, and research items that exceed a 5-minute budget are reclassified as user decisions.