fix(parallel-issues): name the selection helper by its shipped path

[thewrz]

This was written agentically; verify its assertions:

Why

A Codex root running parallel-issues 0.9.12 read bare pick-issues.sh in Step 2b, guessed $agentkit/parallel-issues/scripts/, searched with rg --files (which skips the hidden .shared/), reported the shipped helper missing, and ended the run. That cost one wasted root turn and one human round trip with zero dispatches (#889); this change removes both.

What

• Step 2b of parallel-issues/SKILL.md names the helper by its shipped path where the section first mentions it, and degraded=yes now requires ls -l "$agentkit/.shared/scripts/pick-issues.sh" output plus the failure text, never a guessed path. Wording in the same paragraph was trimmed so the body's estimated tokens fell (66,146 → 66,142 bytes; 16,536 → 16,535 est. tokens). The [parallel-issues] ceiling is untouched.
• triage-issues.sh --help ends its recipe with the next command, "$agentkit/.shared/scripts/pick-issues.sh" --help, so the path reaches the root as tool output. Line count unchanged (576); the helper tree ceiling moves to the exact new measurement (477,429 → 477,442).
• tests/lint-helper-refs.sh holds mention order per heading section: a shared or cross-skill helper named bare before its path in the same section is a violation. A skill's own scripts/ helpers are exempt because the natural guess lands there; a heading-shaped shell comment inside a fence is not a section break. On origin/main (147cd00) the gate fails at parallel-issues/SKILL.md:248 and onboard-repo/SKILL.md:250; both are fixed here with zero body growth (onboard-repo 21,036 → 21,032 bytes). Three fixture cases pin the rule in tests/test-helper-refs.sh.
• The stricter "every bare shared-helper name is a violation" rule was measured and rejected for this PR: it flags 32 lines across all four skills, each fix costs ~26–35 bytes in bodies sitting at their ratchets. The section-first variant without the ordering scope flags 21. Those counts are recorded for a follow-up.

Testing

• tests/run-tests.sh green (ALL GREEN, 141 suites, 0 failures, ~2m29s)
• new lint fails on origin/main (2 violations: parallel-issues:248, onboard-repo:250), passes here (23 documents checked)
• SKILL.md token estimate did not grow (16,536 → 16,535; onboard-repo 5,259 → 5,258)
• shellcheck -x -P SCRIPTDIR -S style clean on all five changed shell files
• CI green

🤖 Co-authored by Claude Fable 5.1. Closes #889.

🤖 Generated with Claude Code

[coderabbitai]

Important

Review skipped

Auto reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

⚙️ Run configuration

Configuration used: Organization UI

Review profile: CHILL

Plan: Team

Run ID: 0b6c5b8f-64e2-4fd4-9039-47667cb3c4be

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

• 🔍 Trigger review

---

_{Comment @coderabbitai help to get the list of available commands.}