interviews

Retro Interviews Skill

Each sprint agent writes their own retro interview before shutting down. Team-lead sends a "Shutdown imminent" message, the agent follows this skill to self-document, then pings team-lead ready. Team-lead then sends shutdown_request.

Output: analysis-reports/retro-session/<YYYY-MM-DD>+<sprint-name>/interviews/<your-role>.md Read by: retro:session Phase 2

---

Agent Self-Documentation (Primary Path)

When you receive a "Shutdown imminent" message from team-lead:

Step 1: Determine your role and question set

Your role	Questions to answer
implementer	C1, C2, C3, D1, D2, D3
qa-validator / reviewer	C1, C2, C3, V1, V2, V3
process-engineer	C1, C2, C3, P1, P2, P3
team-lead	C1, C2, C3, TL1, TL2, TL3
any other role	C1, C2, C3

Step 2: Determine the output path

analysis-reports/retro-session/<YYYY-MM-DD>+<sprint-name>/interviews/<your-role>.md

Use today's date and the sprint name from your task context. If no sprint name is available, use the project name or unnamed-sprint. Create the directory if it does not exist.

Step 3: Write your interview file

Answer all 6 questions (3 common + 3 role-specific) using the formats in interview-templates.md. Answer from your session memory — you lived the session, you do not need to read your own transcript.

Use this file header:

# Retro Interview — <your-role>
Sprint: <sprint-name>
Date: <YYYY-MM-DD>
Agent: <your-name>

Step 4: Confirm to team-lead

After writing the file, reply:

Retro interview complete. Written to analysis-reports/retro-session/<path>/interviews/<role>.md. Ready to shut down.

---

Standard Shutdown Imminent Message (Reference for Team-Lead)

Team-lead sends this before every shutdown_request:

Shutdown imminent. Before we close out:

1. Follow `retro:interviews` — answer your 6 questions (3 common + 3 role-specific for your role)
2. Write your interview to: analysis-reports/retro-session/<YYYY-MM-DD>+<sprint-name>/interviews/<your-role>.md
3. Reply when the file is written and you are ready to shut down.

Team-lead waits for the "ready to shut down" reply, then sends shutdown_request.

---

Verifying Coverage

After all agents have shut down:

ls analysis-reports/retro-session/<YYYY-MM-DD>+<sprint-name>/interviews/

Expected: one .md file per agent type active during the sprint.

retro:session can proceed with partial coverage — it flags missing files as a process gap.

---

Legacy Hook Path (Deprecated)

The subagent-stop-interview.sh SubagentStop hook was the original automated path. It is unreliable — SubagentStop fires at agent idle, not at actual shutdown, and last_assistant_message is often absent from the payload. The team-lead driven approach above replaces it as the primary path.

The hook remains in place as a last-resort fallback for sessions where team-lead fails to send the shutdown imminent message.

---

Integration

• Templates: interview-templates.md (this skill's directory)
• Reads results: retro:session Phase 2
• Sprint shutdown protocol: CLAUDE.md Graceful Shutdown section

---

Obsidian Storage

After each interview file is saved to analysis-reports/, archive it to the Neurons vault. Obsidian is assumed to be running — if the write fails, run obsidian help to diagnose the connection.

Project Slug Resolution

Resolve the project slug in this order:

1. Environment variable — if $OFFICE_PROJECT_NAME is set, slugify and use it
2. Beads database — query the sprint board for a bead with a project-* label: bd list --json | jq -r '.[0].labels[]? | select(startswith("project-")) | ltrimstr("project-")'
3. Ask the user — if neither source yields a value, ask once: "What project is this retrospective for?" and use their answer as the slug

Slugify rule: lowercase, spaces → hyphens, remove all special characters except hyphens. Example: "Same Page Preview" → same-page-preview

Resolve the project slug once per sprint (not once per agent). If collecting interviews across multiple agents in the same sprint, reuse the resolved slug.

Sprint Slug

Derived from the sprint folder name already established (the <sprint-name> segment of analysis-reports/retro-session/<YYYY-MM-DD>+<sprint-name>/). Slugify the same way.

Storage Script (per agent interview)

# Resolve project slug (run once per sprint, reuse across agents)
if [ -n "$OFFICE_PROJECT_NAME" ]; then
  PROJECT_SLUG=$(echo "$OFFICE_PROJECT_NAME" | tr '[:upper:]' '[:lower:]' | tr ' ' '-' | tr -cd '[:alnum:]-')
else
  PROJECT_SLUG=$(bd list --json 2>/dev/null | jq -r '.[0].labels[]? | select(startswith("project-")) | ltrimstr("project-")' | head -1)
fi

# If still unset, ask the user (done interactively — not in this script block)
# USER_INPUT captured via AskUserQuestion: "What project is this retrospective for?"
# PROJECT_SLUG=$(echo "$USER_INPUT" | tr '[:upper:]' '[:lower:]' | tr ' ' '-' | tr -cd '[:alnum:]-')

SPRINT_SLUG="<sprint-slug-from-session>"  # e.g. sprint-1, jquery-fixes
AGENT_ROLE="<agent-role>"                 # e.g. implementer, reviewer, process-engineer
DATE=$(date +%Y-%m-%d)
VAULT_PATH="Retrospectives/${DATE}+${PROJECT_SLUG}+${SPRINT_SLUG}/interviews/${AGENT_ROLE}.md"

VAULT_ROOT="$HOME/Vaults/${OBSIDIAN_VAULT_NAME:-Neurons}"
mkdir -p "$VAULT_ROOT/$(dirname "$VAULT_PATH")"
cat > "$VAULT_ROOT/$VAULT_PATH" << 'EOF'
<interview-file-content>
EOF

Vault Document Format

Each interview file stored at Retrospectives/<YYYY-MM-DD>+<project-slug>+<sprint-slug>/interviews/<agent-role>.md must begin with this YAML frontmatter block:

---
project: <project-slug>
sprint: <sprint-slug>
agent_role: <agent-role>
date: <YYYY-MM-DD>
tags: [retro, interview]
---

The frontmatter project: field enables cross-project Obsidian queries:

• tag:interview — see all agent interviews across all projects
• project: same-page-preview — see one project's agent interview history
• agent_role: implementer — see all implementer interviews across all projects and sprints

Project isolation guarantee: Because project-slug is embedded in both the vault path and the project: frontmatter field, interviews for Project A will never surface as context for Project B in a filtered query.