[product-manager-impersonator] Configurable prompt-audit artifact directory

[github-actions]

Feature Idea

Summary: Add a configurable artifact/output directory for prompt-audit helper scripts and workflow instructions instead of hardcoding /tmp paths.

Why a Customer Would Want This

Teams running these workflows in locked-down CI runners or local environments often have restricted or ephemeral /tmp semantics, which makes prompt-audit setup brittle and harder to reproduce. A configurable path improves reliability and makes investigation artifacts easier to retain.

Rough Implementation Sketch

• Add a shared output-root input/env (for example artifact-root or PROMPT_AUDIT_DIR) and plumb it through prompt-audit setup commands.
• Update scripts/extract-lockfile-prompts.sh to default to that configured location rather than /tmp/prompt-audit.
• Update scripts/fetch-workflow-logs.py default output dir from /tmp/gh-aw/logs to the same configurable root (with CLI override preserved).
• Update trigger-prompt-audit.yml instructions to reference the configured directory path consistently.

Why It Won't Be That Hard

The change is localized to two helper scripts plus one trigger workflow prompt/instructions. All needed extension points already exist (--output-dir in Python script and positional output-dir argument in shell script), so this is mostly wiring defaults and docs.

Evidence

• scripts/fetch-workflow-logs.py hardcodes /tmp/gh-aw/logs as default in both usage text and argparse defaults (L14, L140-L141).
• scripts/extract-lockfile-prompts.sh hardcodes /tmp/prompt-audit as the default output directory (L6, L15).
• .github/workflows/trigger-prompt-audit.yml assumes extracted prompts are written to /tmp/prompt-audit/ in assignment instructions (L24, L28).

Duplicate Check

Searched open issues for equivalent requests and did not find an existing match:

• repo:elastic/ai-github-actions is:issue is:open "prompt-audit" "output-dir"
• repo:elastic/ai-github-actions is:issue is:open "fetch-workflow-logs.py" "/tmp/gh-aw/logs"
• repo:elastic/ai-github-actions is:issue is:open "extract-lockfile-prompts.sh" "/tmp/prompt-audit"

Closest related issue is #987 (fail-closed dedupe mode for Scheduled Audit), which changes dedupe behavior and not artifact path configurability.

Note

🔒 Integrity filter blocked 69 items

The following items were blocked because they don't meet the GitHub integrity level.

• #937 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #851 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #850 search_pull_requests: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• #359 search_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Create issue when merged PRs had unresolved comments #875 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Add PR Docs Review workflow #704 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Add target-repo support to scheduled workflows #694 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Shorter replies from pr-actions-detective #684 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Detector / Fixer chaining workaround #676 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Let's add workflows for refactor and new features #359 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Make the close older issue option for scheduled reports configurable #354 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Add deep research workflow #336 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Update PR Footer behavior when upstream features merge #191 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Auto-address PR Feedback #321 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Downstream Updates Needed workflow #245 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• Create a Central Repo Ops Workflow for Action Distribution #248 list_issues: has lower integrity than agent requires. The agent cannot read data with integrity below "approved".
• ... and 53 more items

To allow these resources, lower min-integrity in your GitHub frontmatter:

tools:
  github:
    min-integrity: approved  # merged | approved | unapproved | none

---

What is this? | From workflow: Trigger Product Manager Impersonator

Give us feedback! React with 🚀 if perfect, 👍 if helpful, 👎 if not.