Skip to content
cccskills
FreePublished on GitHub

kb-ingest-batch

Drive agent-knowledge-updater over a batch of staged files in library/raw/. Tracks progress in .batch-progress.json for resume support. Sequential by default; --parallel <N> opt-in (max 5). Single shelf-index rebuild and one consolidated log.md entry at the end.

The English summary is being prepared. Showing the original description.

See how to install

Files included (1)

  • SKILL.md4.8 KB

SKILL.md (original)

Review the instructions the agent will receive before you install.

Deprecated (v0.3.0+): Prefer /sdlc-knowledge-base:kb-ingest-bulk, which adds a parallel map-reduce path that can update existing shared files (this skill is create-only). kb-ingest-batch remains functional for simple create-only batches.

Batch Ingestion

Drive agent-knowledge-updater over every staged file in library/raw/, with progress tracking and resume support. Second stage of the batch workflow: prepare (kb-prepare-batch) then ingest (this skill).

Arguments

ArgumentDescription
(none)Process all .md files in library/raw/ with status: raw
<dir>Process all .md files in this directory
--parallel <N>Dispatch up to N agents concurrently (max: 5)
--retry-failedRe-queue failed entries from a prior run

Resume behaviour

Progress is tracked in library/raw/.batch-progress.json. On re-invocation:

  • completed files are skipped
  • failed files are left alone unless --retry-failed is passed
  • New status: raw files in raw/ are appended to pending

Preflight

  1. Read CLAUDE.md to resolve library_path, shelf_index_path, log_path
  2. Verify agent-knowledge-updater agent is available

Steps

1. Discover files and build/update manifest

python3 -c "
import sys, os, importlib.util, json
PLUGIN_ROOT = os.environ.get('CLAUDE_PLUGIN_ROOT', '')
SCRIPTS = os.path.join(PLUGIN_ROOT, 'scripts')
INIT = os.path.join(SCRIPTS, '__init__.py')
if os.path.isfile(INIT) and 'sdlc_knowledge_base_scripts' not in sys.modules:
    spec = importlib.util.spec_from_file_location(
        'sdlc_knowledge_base_scripts', INIT, submodule_search_locations=[SCRIPTS])
    if spec and spec.loader:
        mod = importlib.util.module_from_spec(spec)
        sys.modules['sdlc_knowledge_base_scripts'] = mod
        spec.loader.exec_module(mod)
from sdlc_knowledge_base_scripts.kb_ingest_batch import (
    discover_raw_files, load_manifest, build_manifest, save_manifest, retry_failed
)
from pathlib import Path
raw_dir = Path('<raw_dir>')
manifest_path = raw_dir / '.batch-progress.json'
existing = load_manifest(manifest_path)
if existing and <retry_failed_flag>:
    existing = retry_failed(existing)
source_files = discover_raw_files(raw_dir)
manifest = build_manifest(source_files, existing=existing)
save_manifest(manifest_path, manifest)
print(json.dumps({'pending': len(manifest['pending']), 'total': manifest['total']}))
"

Replace <raw_dir> and <retry_failed_flag> with resolved values.

2. Process pending files (sequential)

For each file in pending:

a. Build the dispatch prompt using format_batch_dispatch_prompt():

BATCH_MODE: create-only

Integrate the following source into the knowledge base. Batch mode constraints:
(1) Do NOT modify existing library files — record conflict-existing-file status and stop;
(2) Do NOT run kb-rebuild-indexes;
(3) Do NOT append to log.md.

Source: <path>
Library: <library_path>
Shelf-index: <shelf_index_path>

b. Dispatch agent-knowledge-updater with this prompt using the Agent tool.

c. Update manifest after each dispatch (atomic write):

  • Success or conflict-existing-file → mark_completed()
  • Error or timeout → mark_failed()

d. Parallel mode (--parallel <N>): dispatch up to N agents concurrently using parallel Agent tool calls. Update manifest after each group.

3. Final phase

After all dispatches complete:

a. Rebuild shelf-index (one run):

python3 -c "... from sdlc_knowledge_base_scripts.build_shelf_index import main; sys.exit(main(['<library_path>']))"

If the rebuild exits with code 2, the bulk-add safety rail refused it (>= 50 entries added and index >= 2x; directories can be excluded via <library>/.kb-index-ignore). Neither _shelf-index.md nor log.md was written. You MUST show the user the per-directory breakdown and re-run with --force only after the user confirms — never automatically. Do not write the log entry below until the rebuild has succeeded.

b. Write consolidated log.md entry:

## [YYYY-MM-DD] ingest-batch | <total>/<succeeded>/<failed>

c. Print summary table.

BATCH_MODE constraints on agent-knowledge-updater

Every dispatch includes BATCH_MODE: create-only. The agent:

  1. Does NOT modify existing library files (records conflict-existing-file)
  2. Does NOT run kb-rebuild-indexes
  3. Does NOT append to log.md

Run /sdlc-knowledge-base:kb-ingest individually for conflict files.

Reviews

No reviews yet. Tell others what you thought.

Comments

No login needed. Share how it worked for you, ask a question, or note where it did not run.

No comments yet. Be the first to write one.

Comments are published right away. Do not include personal information such as email addresses or phone numbers. Up to 2 URLs.

0 / 1,000 characters

This form is protected by reCAPTCHA, and the Google Privacy Policy and Terms of Service apply.

More skills from this repository

Generate a GitHub Actions workflow for Android — JDK + Gradle caching, build, unit + instrumented tests, Android Lint, run the Android pre-flight checks, and (optionally) sign + upload a signed app bundle to a Play track. Use when setting up CI for an Android app.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

Validated Google Play release — runs the pre-flight checks and clears the Play policy gates, builds a signed app bundle with a git-stamped versionCode, uploads to a track, and walks staged rollout (with halt/roll-forward). Use when releasing an Android app to Play.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

Scaffold a new Android app with SDLC and Play-safe defaults — Kotlin/Compose Gradle project with version catalog, convention plugins, module layout, a targetSdk that meets the Play mandate, R8 on release, and git-ignored signing so it passes the pre-flight checker from day one. Use when starting a new Android project.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

Set up Android app signing safely — generate an upload keystore, enroll in Play App Signing, and wire a Gradle release signingConfig that reads secrets from a git-ignored keystore.properties / CI (never committed). Use when configuring signing for a Play release.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

Pick an existing workflow for a task or author a new one — recommends workflow + team formation, or generates a fresh workflow YAML with command prompts and team validation.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

Guide the team through writing a change-impact record for IEC 62304 / FDA / ISO 26262 ASIL C+ projects. Surfaces the affected REQ/DES/TEST/CODE artefacts and captures human-attested verification approach.

The English summary is being prepared. Showing the original description.

SteveGJones/ai-first-sdlc-practices41Updated Oct 5, 2026

See all skills by SteveGJones

Report a problem with this skill