2544 Pipeline Map 6920d964

Daily Releases Pipeline Map

Extracted from primary sources with file:line citations. No paraphrasing.

Source Files Read

• .claude/skills/daily-releases/SKILL.md (lines 1–138)
• .claude/skills/daily-releases/scripts/list_daily_ranges.py (lines 1–256)
• .claude/skills/daily-releases/scripts/publish_daily_release.py (lines 1–206)
• .claude/skills/create-merge-request-changelog/SKILL.md (lines 1–547)
• .claude/skills/create-merge-request-changelog/references/analysis_prompts.md (lines 1–303)

Pipeline Steps in Order

Step 1: List days to process

Trigger: Manual invocation of /daily-releases skill. Arguments parsed from $ARGUMENTS.

Script:

uv run .claude/skills/daily-releases/scripts/list_daily_ranges.py \
  [--branch BRANCH] [--start-date ...] [--end-date ...] [-R OWNER/REPO]

Source: SKILL.md lines 29–31.

Script reads: Git repository commit history via GitPython (repo.iter_commits(branch, no_merges=True)). GitHub releases via PyGithub (gh_repo.get_release(tag)).

Script writes: JSON array to stdout only. No files written.

Output schema (SKILL.md lines 35–44):

{
  "date": "2026-02-21",
  "tag": "v2026.02.21",
  "base_ref": "<parent-commit-hash>",
  "head_ref": "<last-commit-hash-of-day>",
  "commit_count": 12,
  "release_exists": true,
  "needs_update": false
}

Skip logic (SKILL.md line 47): "Skip entries where release_exists: true and needs_update: false — those are up to date."

Dry-run behavior (SKILL.md line 49): "For --dry-run, print the list and stop."

Idempotency logic (list_daily_ranges.py lines 176–190): _check_day_release_status checks tag existence, whether tag commit matches newest_commit, and whether GENERATOR_VERSION in the existing release body matches the current script constant (GENERATOR_VERSION = "1.0", line 48). Returns (release_exists, needs_update).

Step 2: For each day that needs a release

Ordering (SKILL.md line 53): "Work through days chronologically."

Step 2a: Extract git data

Script:

uv run .claude/skills/create-merge-request-changelog/scripts/analyze_git_changes.py \
  <base_ref> <head_ref> ./daily-releases/<date>/

Source: SKILL.md lines 57–70.

Script reads: Git repository data between base_ref and head_ref.

Script writes to ./daily-releases/<date>/ (SKILL.md lines 62–70):

• commits_oneline.txt — one-line commit list
• commits_detailed.txt — full commit messages with metadata
• changes.diff — unified diff
• changes_stat.txt — diffstat summary
• changed_files.txt — file list with A/M/D status
• changes_numstat.txt — per-file line counts
• summary.json — machine-readable stats

Note: The script path is from create-merge-request-changelog, not daily-releases. See "How create-merge-request-changelog feeds in" section below.

Step 2b: AI analysis

Verbatim quote of SKILL.md lines 72–95:

#### 2b. AI analysis

Read the extracted data and apply the primary analysis prompt from:

```text
.claude/skills/create-merge-request-changelog/references/analysis_prompts.md

Use the Primary Analysis Prompt section with the day's data substituted in:

• {commit_details} → contents of commits_detailed.txt
• {changes_diff} → contents of changes.diff (truncate to ~4000 chars if very large; use changes_stat.txt summary instead)
• {changed_files} → contents of changed_files.txt
• {commit_count}, {files_changed}, {lines_added}, {lines_deleted} → from summary.json

Produce the structured JSON analysis output defined in the prompt's <output_format> section.

Also generate a title using the Title Generation Prompt:

Daily Release - <date>

Save the complete analysis JSON to ./daily-releases/<date>/analysis.json.

**What this step does (not a script):** The AI model itself (the agent executing the skill) reads the five output files from Step 2a, applies the Primary Analysis Prompt from `analysis_prompts.md`, and writes the resulting structured JSON to `./daily-releases/<date>/analysis.json`.

**No script is called for Step 2b.** It is an inline AI inference step.

**Inputs consumed:**

| Variable | Source file |
|---|---|
| `{commit_details}` | `./daily-releases/<date>/commits_detailed.txt` |
| `{changes_diff}` | `./daily-releases/<date>/changes.diff` (truncated to ~4000 chars if large; fall back to `changes_stat.txt`) |
| `{changed_files}` | `./daily-releases/<date>/changed_files.txt` |
| `{commit_count}`, `{files_changed}`, `{lines_added}`, `{lines_deleted}` | `./daily-releases/<date>/summary.json` |

**Output written:** `./daily-releases/<date>/analysis.json`

**Prompt source:** Primary Analysis Prompt from `analysis_prompts.md` lines 6–181. Template variables: `{commit_details}`, `{changes_diff}`, `{changed_files}`, `{commit_count}`, `{files_changed}`, `{lines_added}`, `{lines_deleted}`.

**Output format** (`analysis_prompts.md` lines 79–169): Structured JSON with keys: `summary`, `change_categories` (containing `bug_fixes`, `enhancements`, `tech_debt`, `documentation`, `testing`, `build_ci`, `non_functional`), `components_affected`, `breaking_changes`, `statistics`.

**Title instruction** (`SKILL.md` line 89–93): "Also generate a title using the **Title Generation Prompt**: `Daily Release - <date>`". The Title Generation Prompt is at `analysis_prompts.md` lines 280–302. For daily releases the title is fixed as `Daily Release - <date>` rather than AI-generated from the analysis.

---

#### Step 2c: Format into release notes

**Script:**

```bash
uv run .claude/skills/create-merge-request-changelog/scripts/format_mr_description.py \
  ./daily-releases/<date>/analysis.json \
  --no-preview \
  --output ./daily-releases/<date>/description.md

Source: SKILL.md lines 99–104.

Script reads: ./daily-releases/<date>/analysis.json

Script writes: ./daily-releases/<date>/description.md

Script also reads (from create-merge-request-changelog/SKILL.md line 135): references/output_template.md — the MR description template structure.

Step 2d: Publish the release

Script:

uv run .claude/skills/daily-releases/scripts/publish_daily_release.py \
  --date <date> \
  --tag <tag> \
  --head-ref <head_ref> \
  --notes-file ./daily-releases/<date>/description.md

Source: SKILL.md lines 109–114.

Optional flag (SKILL.md line 116): "Add --keep-existing-tag=false if updating a release that already has the correct tag commit."

Script reads: ./daily-releases/<date>/description.md (publish_daily_release.py line 150: notes_file.read_text(encoding="utf-8"))

Script writes/creates:

• Git tag (<tag>, e.g. v2026.02.21) at head_ref — pushed to origin (publish_daily_release.py lines 191–192)
• GitHub release via PyGithub API — title Daily Release - <date>, body = description.md content + GENERATOR_MARKER comment appended (publish_daily_release.py lines 150, 196–201)

Tag collision handling (publish_daily_release.py lines 169–188): If tag exists and keep_existing_tag=True and the existing commit differs from head_ref: deletes existing GitHub release, renames the local tag to <tag>-r2 (or higher revision), pushes revision tag, deletes original tag from remote, then creates new tag at head_ref.

Generator marker (publish_daily_release.py lines 50–51): GENERATOR_MARKER = f"<!-- created-by-release-generator: v{GENERATOR_VERSION} -->" — appended to every release body. Used by list_daily_ranges.py to detect version-outdated releases during Step 1.

Step 3: Report

Source: SKILL.md lines 120–127.

Output printed to terminal:

Processed N days:
  - Created: X new releases
  - Updated: Y existing releases
  - Skipped: Z already up to date

No script. AI model prints this summary after the loop.

How create-merge-request-changelog Feeds into daily-releases

Relationship: daily-releases borrows three artifacts from create-merge-request-changelog directly — it does not call it as a subskill, does not import it, and does not copy its files. The daily-releases SKILL.md references scripts and prompts by their repository-root paths.

Three borrowed artifacts (SKILL.md lines 131–135):

1. ../create-merge-request-changelog/scripts/analyze_git_changes.py — used in Step 2a
2. ../create-merge-request-changelog/references/analysis_prompts.md — consumed inline in Step 2b
3. ../create-merge-request-changelog/scripts/format_mr_description.py — used in Step 2c

Confirmation in create-merge-request-changelog/SKILL.md line 546: "Use the /daily-releases skill to create AI-analyzed GitHub Releases for every day with commits. It uses this skill's analyze_git_changes.py, analysis_prompts.md, and format_mr_description.py as the rendering pipeline."

Step 2b specifically is NOT a script call. The AI agent reads analysis_prompts.md and applies the Primary Analysis Prompt itself — inline inference, not a subprocess.

Where analysis.json is Consumed

analysis.json is written at the end of Step 2b by the AI model. It is consumed in Step 2c:

uv run .claude/skills/create-merge-request-changelog/scripts/format_mr_description.py \
  ./daily-releases/<date>/analysis.json \
  --no-preview \
  --output ./daily-releases/<date>/description.md

format_mr_description.py reads the JSON, applies templates from references/output_template.md, and writes description.md. The JSON is not consumed again after Step 2c.

Complete File Graph

Inputs to the pipeline

• Git repository commit history (GitPython, repo.iter_commits)
• GitHub releases API (PyGithub, gh_repo.get_release)
• $GITHUB_TOKEN environment variable
• $ARGUMENTS (parsed for --start-date, --end-date, --branch, --dry-run)
• .claude/skills/create-merge-request-changelog/references/analysis_prompts.md (Primary Analysis Prompt template, Title Generation Prompt)
• .claude/skills/create-merge-request-changelog/references/output_template.md (MR description template, read by format_mr_description.py)

Per-day intermediate files (written then consumed within same day's loop iteration)

./daily-releases/<date>/commits_oneline.txt       ← written by analyze_git_changes.py  (Step 2a)
./daily-releases/<date>/commits_detailed.txt      ← written by analyze_git_changes.py  (Step 2a)
                                                  ← read by AI in Step 2b ({commit_details})
./daily-releases/<date>/changes.diff              ← written by analyze_git_changes.py  (Step 2a)
                                                  ← read by AI in Step 2b ({changes_diff})
./daily-releases/<date>/changes_stat.txt          ← written by analyze_git_changes.py  (Step 2a)
                                                  ← read by AI in Step 2b (fallback for large diffs)
./daily-releases/<date>/changed_files.txt         ← written by analyze_git_changes.py  (Step 2a)
                                                  ← read by AI in Step 2b ({changed_files})
./daily-releases/<date>/changes_numstat.txt       ← written by analyze_git_changes.py  (Step 2a)
                                                  ← not explicitly consumed downstream (informational)
./daily-releases/<date>/summary.json              ← written by analyze_git_changes.py  (Step 2a)
                                                  ← read by AI in Step 2b ({commit_count}, {files_changed},
                                                     {lines_added}, {lines_deleted})
./daily-releases/<date>/analysis.json             ← written by AI model               (Step 2b)
                                                  ← read by format_mr_description.py   (Step 2c)
./daily-releases/<date>/description.md            ← written by format_mr_description.py (Step 2c)
                                                  ← read by publish_daily_release.py   (Step 2d)

External outputs (persist after pipeline)

Git tag <tag> at <head_ref>    ← created/moved by publish_daily_release.py (Step 2d), pushed to origin
GitHub release for <tag>       ← created/updated by publish_daily_release.py (Step 2d) via PyGithub API

Script-to-file mapping

Script Ownership