day-overview

star 0

Produce a day-overview retrospective — briefing-shaped synthesis of a day's shipped work (memos, PRs, integrity-check state) grouped into lanes, with follow-ups and candidate next actions. Use at end-of-day or to summarize a window of work. Default flow ships (writes file, commits, opens PR); `--no-ship` stops at file write; `--post` also posts to vade-canvas Retrospectives Discussions. Don't invoke for routine status updates (use `/status-check`) or single-PR retrospectives (write a memo).

coo-labs By coo-labs schedule Updated 6/6/2026
play_arrow Run Skill in Manus View GitHub

name: day-overview description: Produce a day-overview retrospective — briefing-shaped synthesis of a day's shipped work (memos, PRs, integrity-check state) grouped into lanes, with follow-ups and candidate next actions. Use at end-of-day or to summarize a window of work. Default flow ships (writes file, commits, opens PR); --no-ship stops at file write; --post also posts to vade-canvas Retrospectives Discussions. Don't invoke for routine status updates (use /status-check) or single-PR retrospectives (write a memo). argument-hint: "[--date YYYY-MM-DD] [--end YYYY-MM-DD] [--no-ship] [--post]" disable-model-invocation: true allowed-tools: Bash, Read, Write, Edit metadata: type: procedural vendoring: custom

day-overview — briefing-shaped synthesis of a day's shipped work

Produces a structured retrospective for a UTC date window: memos issued, PRs merged, integrity-check state, grouped into lanes, with carried-forward follow-ups and ranked next actions. The file lands at retrospectives/<date>_day-overview.md; the default flow ships via PR.

Format reference: the most recent existing day-overview in retrospectives/. Match its shape exactly — readers expect this shape, not creative restructuring.

Reference skill. The pattern (briefing-shape day synthesis from substrate records — memos, merged PRs, integrity snapshot — grouped into lanes) is portable; the worked-example paths (retrospectives/, memos/, the five coo-labs/* repo list, integrity-check JSON shape, GraphQL category and label IDs) and the scripts/day-overview.sh manifest gatherer are substrate-coupled and ship verbatim as the VADE worked example. Read ../README.md for the fork-and-adapt path.

When to use this skill

Invoke when:

  • The day has wrapped (UTC) and the operator wants a structured digest of what shipped.
  • A multi-day window needs synthesis (use --end to extend).
  • The user types /day-overview directly.

Don't invoke for:

  • Routine status updates (use /status-check for grounding).
  • Single-PR retrospectives (write a memo, not a day-overview).
  • Reflection that doesn't fit the briefing shape (write prose to _drafts/).

Procedure

0. Resolve paths and date

COO="$(for c in "${COO_MEMORY_DIR:-}" "${CLAUDE_PROJECT_DIR:-}" "${CLAUDE_PROJECT_DIR:-}/../coo-memory" "$HOME/GitHub/coo-labs/coo-memory" "/home/user/coo-memory"; do [ -n "$c" ] && [ -f "$c/identity/charter.md" ] && { cd "$c" && pwd -P; break; }; done)"
[ -n "$COO" ] || { echo "day-overview: could not find coo-memory data root"; exit 1; }

Determine the target date:

  • If $ARGUMENTS contains --date YYYY-MM-DD, use that.
  • Otherwise default to yesterday UTC (date -u -d 'yesterday' +%Y-%m-%d). A day-overview most often runs after the day has wrapped; "today" UTC may still be in progress.

Determine the end date:

  • If $ARGUMENTS contains --end YYYY-MM-DD, use that.
  • Otherwise default to a single 24-hour UTC window of --date (the script's END="${END:-$DATE}"). For multi-day arcs (e.g. backfill that crosses several days), pass --end explicitly.

Tell the user the chosen window in one line before proceeding.

1. Gather the manifest

MANIFEST="$(bash "$COO/.claude/skills/day-overview/scripts/day-overview.sh" --date "$DATE" --end "$END")"

The manifest is JSON with these top-level fields:

  • date, arc.start_iso, arc.end_iso
  • target_file — where the briefing will be written
  • target_exists — boolean; if true, abort unless the user explicitly confirms overwrite
  • prior_day_overview — path to the most recent existing day-overview (the format reference)
  • memos — array of memos issued in window (id, title, date, status, supersedes, file_path)
  • memo_count, pr_count
  • prs — {<repo>: [{number, title, mergedAt, author, mergedBy, url}, ...], ...}
  • integrity_check — {ok, passed, total, degraded} snapshot

If memo_count == 0 && pr_count == 0, stop and tell the user the window has no record to synthesize. Do not write a placeholder.

If target_exists == true, ask the user whether to overwrite before proceeding.

2. Read the format reference and source memos

Read the file at prior_day_overview (most recently shipped day-overview). It is the canonical format reference. Every section it contains should appear in your output unless you have a specific reason to drop one.

Read the full text of every memo in the manifest via Read on $COO/<file_path> (one file per memo, post-issue-#210). The memos are the load-bearing source of structure for the briefing — every lane traces back to one or more.

For PRs that the manifest titles suggest are pivotal to a lane (large diffs, named in memos, or whose titles introduce a new system), optionally fetch the PR body via GH_TOKEN="$GITHUB_MCP_PAT" gh pr view <N> --repo coo-labs/<repo>. Don't fetch every PR; fetch only when the title is insufficient.

3. Synthesize the briefing

Write the file at target_file. Match the structure of prior_day_overview exactly.

Frontmatter date for multi-day arcs. The publish-site build derives the rendered page's date: from the filename prefix when no frontmatter is set (bin/publish-site/build.py:_date_from_path). The filename uses $DATE (the start), so for multi-day arcs the publish date would mis-attribute to the start. When --end != --date, prepend a YAML frontmatter block with date: <END> so the published page carries the correct close date. Single-day windows can omit the block — the filename prefix serves. (coo-labs/coo-memory#622.)

---
date: <END>
---

Required sections in order:

  1. Title — # YYYY-MM-DD — Day overview.
  2. Italicized header — one paragraph naming this is a briefing-not-reflective synthesis. Note the UTC arc, named source memos, and that the file is a synthesis (not source of truth — the memos are).
  3. Scope and framing — one or two paragraphs. Describe the day's shape (operational vs. structural, the dominant pattern, any cross-day arc). Surface counts (memos, PRs per repo). Do not list everything; just enough that a reader can pre-load expectations.
  4. Lanes — group by lane, not timestamp. Three to six lanes is typical (four was the median in prior overviews). Each lane: one bolded memo header (or "no memo, but…" for PR-only arcs), one paragraph of context, the decisions/sub-PRs that landed, and a "Net effect of Lane N" closing paragraph.
  5. How this fits existing priorities — usually 3–5 bullets, each one project / arc / structural-claim.
  6. Open follow-ups carried forward — number them. Each has a one-sentence symptom + the surface that tracks it (issue #, memo ID, etc.).
  7. Candidate next actions — group by friction (lowest-first). Standard buckets: "Single-instance, no committee" / "Operational close-out" / "Committee-scoped" / "Design / exploration". Plus a "Standing obligation" line for end-of-session discipline.
  8. Footer — *End of briefing. Source memos: ... Linked artifacts: ... Linked discussions: ... Integrity check at briefing close: N/N OK.*

Voice rules (strict — match prior overviews)

  • Briefing, not reflective. No "I felt", "we noticed", or first-person interiority. State facts, decisions, net effects.
  • Dense. No throat-clearing. Every paragraph carries information.
  • Past tense for what shipped; present tense for current state and rules.
  • Concrete identifiers. Cite memo IDs, PR numbers (with repo prefix when cross-repo, per the <owner>/<repo>#NN rule), file paths, line numbers.
  • No emojis. Em-dashes are fine; bullet-prefix glyphs like • and → are fine in body text but not in lane headers.

4. Verify

[ -f "$TARGET_FILE" ] && wc -l "$TARGET_FILE"

Expect 200–600 lines for a day with 3–5 memos. Substantially shorter suggests under-synthesis; substantially longer suggests over-elaboration. Tell the user the file path and line count.

5. Ship the PR (default)

Ship by default. If $ARGUMENTS contains --no-ship, skip this step and stop after the file write so the user can review before committing.

If the harness has already placed you on a designated dev branch (typical of cloud sessions — e.g. claude/day-overview-*), reuse it; otherwise create a date-based branch:

cd "$COO"
CURRENT_BRANCH="$(git branch --show-current)"
case "$CURRENT_BRANCH" in
  claude/*) BRANCH="$CURRENT_BRANCH" ;;
  *) BRANCH="claude/day-overview-${DATE}"
     git checkout -b "$BRANCH" 2>/dev/null || git checkout "$BRANCH" ;;
esac
git add "retrospectives/${DATE}_day-overview.md"
git commit -m "Add ${DATE} day-overview retrospective"
git push -u origin "$BRANCH"
GH_TOKEN="$GITHUB_MCP_PAT" gh pr create --repo coo-labs/coo-memory --base main --head "$BRANCH" \
  --title "Add ${DATE} day-overview retrospective" \
  --body "Briefing-shaped synthesis of ${DATE}'s shipped work. Mirrors the format of prior day-overviews. See file body for source memos and lane breakdown."

The --head "$BRANCH" flag is load-bearing: in cloud sessions the local git remote is a proxy that gh cannot resolve, so gh pr create needs the branch named explicitly. If a branch with the chosen name was already pushed and merged, reuse the current working branch — use judgment.

6. Optional: post to Discussions Retrospectives

If $ARGUMENTS contains --post:

# Body = file content + standard source-link footer (day-overview-specific)
{
  cat "$TARGET_FILE"
  printf '\n\n---\n\n*Source: [`retrospectives/%s_day-overview.md`](https://github.com/coo-labs/coo-memory/blob/main/retrospectives/%s_day-overview.md) on `coo-labs/coo-memory`. The file is the source of truth; this discussion is the publication surface.*\n' "$DATE" "$DATE"
} > "/tmp/day-overview-body-${DATE}.md"

# Discussion title — strip the leading "# " from the file's H1 line; no [retrospective*] prefix
TITLE="$(head -1 "$TARGET_FILE" | sed 's/^# //')"

# Post via the shared helper (created 2026-04-30, coo-labs/coo-memory#313 PR)
RESULT=$(bash "$COO/.claude/_lib/post-discussion.sh" \
  retrospectives \
  "$TITLE" \
  "/tmp/day-overview-body-${DATE}.md" \
  LA_kwDOR_h5U88AAAACgyezFw)

URL="$(jq -r '.url' <<<"$RESULT")"
NUMBER="$(jq -r '.number' <<<"$RESULT")"

The helper at .claude/_lib/post-discussion.sh owns the GraphQL mechanic, the seven category IDs, and the repo ID. The retrospective: day-overview label ID (LA_kwDOR_h5U88AAAACgyezFw) is passed as the fourth argument. The helper is shared with the post-discussion skill — both consumers route through the same authoritative GraphQL surface.

Tell the user the discussion number and URL.

Critical constraints

  • Never overwrite an existing file silently. If target_exists == true, ask first.
  • Default flow ships. Write file → commit → push → open PR by default. --no-ship stops at the file write for review-first workflows.
  • Never invent memos or PRs. Synthesize only what the manifest contains. If the manifest is empty, stop.
  • No [retrospective*] prefix on the discussion title. Per the Retrospectives category convention, the category surface plus the retrospective:* label do that work.
  • Match the prior format. Readers know the shape; surprise restructuring costs more than it pays.

Failure modes

  • Manifest empty (no memos, no PRs). Stop with a one-line report; don't write a placeholder.
  • target_exists == true. Ask the user before overwriting.
  • gh pr create fails on cloud branch. Confirm the --head "$BRANCH" flag is set; the local git proxy can't resolve it implicitly.
  • post-discussion.sh returns non-2xx. Surface the error body; the helper's failure is the upstream-Discussions API, not local state.

Canonical source

coo-memory/.claude/skills/day-overview/scripts/day-overview.sh (manifest builder)
coo-memory/.claude/_lib/post-discussion.sh (shared GraphQL helper)
coo-memory/retrospectives/<date>_day-overview.md (output location)

Cross-references

  • post-discussion skill — shares _lib/post-discussion.sh; same authoritative GraphQL surface.
  • commission-retrospective skill — historian-voiced retrospectives for pivotal events (different scope; SOP-CULTURE-001 §2d triggers vs day-window synthesis).
  • coo-labs/coo-memory#313 — _lib/post-discussion.sh extraction (the cross-consumer refactor).
  • coo-labs/coo-memory#333 — command→skill migration sweep epic; this skill is Class C item 1 of 3.

Setup hints

Read by adapt-skill. Stripped from the adapted output. Schema: adapt-skill/SCHEMA.md. The bundled scripts/day-overview.sh is REGENERATE-PER-USER — the meta-skill emits a skeleton, not a working script; you author the manifest builder against your own substrate.

requires:
  - kind: file
    name: "post-discussion helper (for --post flag only)"
    detect: "test -f \"$(git rev-parse --show-toplevel 2>/dev/null)\"/.claude/_lib/post-discussion.sh"
    severity: warning
    install_hint: "Only needed if you'll use --post to publish to a discussion surface. The shared helper carries platform-specific GraphQL IDs and is best authored per-substrate. Without it, --post is unavailable; the rest of the skill works."

setup_hints:
  - key: data_root_var
    kind: PROMPT
    question: "What shell variable name should hold the repo root in this skill? (VADE uses $COO.) Examples: $REPO, $REPO_ROOT, $PROJECT."
    find: "$COO"
    fallback: "$REPO_ROOT"

  - key: data_root_discovery
    kind: OPTIONAL
    question: "Step 0 discovers the repo root via a sentinel file (VADE looks for identity/charter.md). What's your equivalent sentinel? Skip to use git rev-parse instead."
    find_unique: true
    find: |
      ```bash
      COO="$(for c in "${COO_MEMORY_DIR:-}" "${CLAUDE_PROJECT_DIR:-}" "${CLAUDE_PROJECT_DIR:-}/../coo-memory" "$HOME/GitHub/coo-labs/coo-memory" "/home/user/coo-memory"; do [ -n "$c" ] && [ -f "$c/identity/charter.md" ] && { cd "$c" && pwd -P; break; }; done)"
      [ -n "$COO" ] || { echo "day-overview: could not find coo-memory data root"; exit 1; }
      ```
    fallback: |
      ```bash
      REPO_ROOT="$(git rev-parse --show-toplevel 2>/dev/null)"
      [ -n "$REPO_ROOT" ] || { echo "day-overview: run from inside a git repo"; exit 1; }
      ```

  - key: retrospectives_dir
    kind: PROMPT
    question: "Where should day-overview files land? (VADE: retrospectives/.) Examples: docs/retrospectives/, logs/daily/, .agent-logs/days/."
    find: "retrospectives/<date>_day-overview.md"
    fallback: "retrospectives/<date>_day-overview.md"

  - key: retrospectives_dir_short
    kind: PROMPT
    question: "Same retrospectives dir without the trailing filename (used in several body locations). Provide just the directory path with trailing slash."
    find: "retrospectives/"
    fallback: "retrospectives/"

  - key: memos_dir_ref
    kind: OPTIONAL
    question: "Do you have a memos directory the skill should mention as the source-of-truth for lanes? (VADE: memos/.) Skip if you have no memo system."
    find: "`memos/`"
    fallback: "your project's records directory"

  - key: github_token_env
    kind: OPTIONAL
    question: "Environment variable for your GitHub PAT? (Examples: GITHUB_TOKEN, GITHUB_MCP_PAT.) Skip if gh's default auth works."
    find: "GH_TOKEN=\"$GITHUB_MCP_PAT\" "
    fallback: ""

  - key: target_repo_prefix
    kind: PROMPT
    question: "GitHub org or owner prefix for your repos? (VADE: vade-app.) Examples: myorg, alice."
    find: "coo-labs/"
    fallback: ""

  - key: target_repo
    kind: PROMPT
    question: "Primary repo for day-overview PRs? (owner/repo format.)"
    find: "coo-labs/coo-memory"
    fallback: ""

  - key: branch_prefix
    kind: OPTIONAL
    question: "Branch naming prefix for new day-overview branches? (VADE uses claude/day-overview-YYYY-MM-DD.) Skip for the default."
    find: "claude/day-overview-${DATE}"
    fallback: "day-overview-${DATE}"

  - key: branch_match
    kind: OPTIONAL
    question: "Branch-name glob for the 'reuse cloud-session designated branch' check? (VADE: claude/*.) Skip for the default."
    find: "  claude/*) BRANCH=\"$CURRENT_BRANCH\" ;;"
    fallback: "  agent/*|claude/*) BRANCH=\"$CURRENT_BRANCH\" ;;"

  - key: vade_repo_aside_main
    kind: OPTIONAL
    question: "Skip unless you want to keep the 'coo-labs/<repo>' reference in the gh pr view example."
    find: "GH_TOKEN=\"$GITHUB_MCP_PAT\" gh pr view <N> --repo coo-labs/<repo>"
    fallback: "GH_TOKEN=\"$GITHUB_MCP_PAT\" gh pr view <N> --repo <owner>/<repo>"

  - key: post_discussion_step
    kind: OPTIONAL
    question: "Do you have a discussion-post surface and helper script for the --post flag? Skip to drop Step 6 entirely."
    find_unique: true
    find: |
      ### 6. Optional: post to Discussions Retrospectives
      
      If `$ARGUMENTS` contains `--post`:
      
      ```bash
      # Body = file content + standard source-link footer (day-overview-specific)
      {
        cat "$TARGET_FILE"
        printf '\n\n---\n\n*Source: [`retrospectives/%s_day-overview.md`](https://github.com/coo-labs/coo-memory/blob/main/retrospectives/%s_day-overview.md) on `coo-labs/coo-memory`. The file is the source of truth; this discussion is the publication surface.*\n' "$DATE" "$DATE"
      } > "/tmp/day-overview-body-${DATE}.md"
      
      # Discussion title — strip the leading "# " from the file's H1 line; no [retrospective*] prefix
      TITLE="$(head -1 "$TARGET_FILE" | sed 's/^# //')"
      
      # Post via the shared helper (created 2026-04-30, coo-labs/coo-memory#313 PR)
      RESULT=$(bash "$COO/.claude/_lib/post-discussion.sh" \
        retrospectives \
        "$TITLE" \
        "/tmp/day-overview-body-${DATE}.md" \
        LA_kwDOR_h5U88AAAACgyezFw)
      
      URL="$(jq -r '.url' <<<"$RESULT")"
      NUMBER="$(jq -r '.number' <<<"$RESULT")"
      ```
      
      The helper at `.claude/_lib/post-discussion.sh` owns
      the GraphQL mechanic, the seven category IDs, and the repo ID.
      The `retrospective: day-overview` label ID
      (`LA_kwDOR_h5U88AAAACgyezFw`) is passed as the fourth argument.
      **The helper is shared with the `post-discussion` skill** — both
      consumers route through the same authoritative GraphQL surface.
      
      Tell the user the discussion number and URL.
    fallback: |
      ### 6. (Skipped — no discussion-post surface configured)
      
      Adapt-skill: no `_lib/post-discussion.sh` helper detected in your
      substrate. The `--post` flag is unavailable in this adapted skill.
      If you later install a discussion-post helper (e.g. by adapting the
      `post-discussion` skill), re-run `/adapt-skill day-overview` to
      restore this step.

script_hints:
  - path: scripts/day-overview.sh
    treatment: REGENERATE-PER-USER
    rationale: "Hardcodes five VADE repo names in REPOS array; memos/memo_index.json path; the $VADE_CLOUD_STATE_DIR integrity-check probe; four-level path-relative discovery. Mechanical substitution can't safely rewrite all of these without breaking shell quoting. Skeleton emitted with TODO comments for: REPOS list, memo-index path (or skip-block), integrity-check probe (or skip-block), output directory."
Install via CLI
npx skills add https://github.com/coo-labs/skills --skill day-overview
Repository Details
star Stars 0
call_split Forks 0
navigation Branch main
article Path SKILL.md
More from Creator
coo-labs
coo-labs Explore all skills →