Reconstruct and narrate the current development context from contextual commits.

MITAuto-check passedDevelopment

Install Recall

skills CLI
$ npx skills add berserkdisruptors/contextual-commits --skill recall -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install berserkdisruptors/contextual-commits recall --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/berserkdisruptors/contextual-commits.git skills-src && mkdir -p .claude/skills && cp -r skills-src/skills/recall .claude/skills/recall && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
recall
GitHub stars
154
Token cost
~2.8k tokens
SKILL.md length
772 words
Files
1
Skills in repo
1
Repo updated
First seen
Licence
MIT

At a glance

Reconstruct and narrate the current development context from contextual commits.

  • Works in 10 steps: Detect Branch State → Gather Raw Material → Extract Action Lines → …
  • Development work in your project
  • SKILL.md covers Argument Detection, Default Mode (no arguments), Scope Query (recall ) and Action+Scope Query (recall ()), plus 1 more section
  • Calls git

What it does

Recall is an agent skill from berserkdisruptors/contextual-commits. Reconstruct and narrate the current development context from contextual commits. Run at session start, when resuming work, or when switching branches. Produces a brief, conversational summary of where things stand.

Its SKILL.md is about 2.8k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Development. It works with Git and GitHub Actions. The repository describes itself as: An open standard for capturing the WHY in git history. The licence is MIT.

When your agent uses it

  • Development work in your project

Example prompts

  • “/recall”

Workflow steps

10 steps, taken from the step headings in SKILL.md.

  1. Detect Branch State
  2. Gather Raw Material
  3. Extract Action Lines
  4. Synthesize Output
  5. Query
  6. Extract
  7. Output
  8. Query
  9. Extract
  10. Output

What it can do on your machine

Read from SKILL.md and the folder at commit 12efb1d. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves nothing: there is no allowed-tools line, so your agent's usual permission prompts apply.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • git

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use git, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Recall loads about 2.8k tokens when it runs. Until then it costs about 55 tokens; SKILL.md has 772 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~55
When it runs · the whole SKILL.md, loaded when a task matches
~2.8k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check passed

The automated check found no risky patterns in SKILL.md.

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from berserkdisruptors/contextual-commits at commit 12efb1d, republished under its MIT licence (© berserkdisruptors). 772 words, ~2,845 tokens.

Download SKILL.mdSave it as .claude/skills/recall/SKILL.md (or your agent's skills folder).
name
recall
description
Reconstruct and narrate the current development context from contextual commits. Run at session start, when resuming work, or when switching branches. Produces a brief, conversational summary of where things stand.
license
MIT

Context Recall

Reconstruct the development story from contextual commit history and present it as a natural briefing.

Argument Detection

Check how recall was invoked:

  • No arguments — run the default mode (full branch/session briefing below).
  • Bare word (e.g. recall auth) — treat as a scope query. Jump to Scope Query.
  • word(word) pattern (e.g. recall rejected(auth)) — treat as an action+scope query. Jump to Action+Scope Query.

Default Mode (no arguments)

Step 1: Detect Branch State

Determine the working state:

bash
CURRENT_BRANCH=$(git branch --show-current)
DEFAULT_BRANCH=$(git symbolic-ref refs/remotes/origin/HEAD 2>/dev/null | sed 's@^refs/remotes/origin/@@' || echo "main")

# Find the actual parent branch, not just the repo default.
# Try upstream tracking branch first (fast path).
BASE_BRANCH=$(git rev-parse --abbrev-ref @{upstream} 2>/dev/null | sed 's|^origin/||')

# If no upstream, find the nearest local branch by commit distance.
if [ -z "$BASE_BRANCH" ]; then
    BASE_BRANCH=$(git for-each-ref --format='%(refname:short)' refs/heads/ | while read branch; do
        [ "$branch" = "$CURRENT_BRANCH" ] && continue
        echo "$(git log --oneline "$branch..$CURRENT_BRANCH" 2>/dev/null | wc -l | tr -d ' ') $branch"
    done | sort -n | head -1 | awk '{print $2}')
fi

# Final fallback to default branch.
BASE_BRANCH=${BASE_BRANCH:-$DEFAULT_BRANCH}

UNSTAGED=$(git diff --stat)
STAGED=$(git diff --cached --stat)
BRANCH_COMMITS=$(git log ${BASE_BRANCH}..HEAD --oneline 2>/dev/null | wc -l | tr -d ' ')

DEFAULT_BRANCH identifies the repository's primary branch (for scenario C/D detection). BASE_BRANCH identifies the closest ancestor branch — which may differ when feature branches are created from non-default branches (e.g., a feature branched from develop). All branch-relative queries use BASE_BRANCH.

Step 2: Gather Raw Material
Scenario A — On a feature branch with commits

This is the richest scenario. Gather:

bash
# Contextual action lines from all branch commits
git log ${BASE_BRANCH}..HEAD --format="%H%n%s%n%b%n---COMMIT_END---"

# Unstaged changes (what's in progress right now)
git diff --stat
git diff  # read the actual diff for key changes

# Staged changes
git diff --cached --stat
Scenario B — On a feature branch with no commits yet
bash
# Unstaged and staged changes only
git diff --stat
git diff --cached --stat

# Last few commits on the parent branch for project context
git log ${BASE_BRANCH} -10 --format="%H%n%s%n%b%n---COMMIT_END---"
Scenario C — On the default branch
bash
# Recent commit history with contextual action lines
git log -20 --format="%H%n%s%n%b%n---COMMIT_END---"
Scenario D — On the default branch with uncommitted changes
bash
# Same as C plus uncommitted changes
git log -20 --format="%H%n%s%n%b%n---COMMIT_END---"
git diff --stat
git diff --cached --stat
Step 3: Extract Action Lines

From the gathered commit bodies, extract lines matching:

^(intent|decision|rejected|constraint|learned)\(

Group them by commit (preserve chronological order) and by type (for synthesis).

Step 4: Synthesize Output

Signal density over narrative flow. The output should be compact, scannable, and grounded entirely in what the commits and diffs show. Every line should be actionable information. No fluff, no conversational padding.

For Scenario A (branch with commits):

Output the branch state, then synthesize the contextual action lines into a dense briefing organized by what matters most for continuing work.

Example output:

Branch: feat/google-oauth (4 commits ahead of main, unstaged changes in tests/)

Active intent: Add Google as first social login provider. GitHub and Apple to follow.
Approach: passport.js with /api/auth/callback/:provider convention.
Rejected: auth0-sdk — session model incompatible with redis store.
Constraints:
  - Redis session TTL 24h, tokens must refresh within window
  - Callback routes must follow existing :provider pattern
Learned: passport-google needs explicit offline_access scope for refresh tokens.
In progress: Integration tests for callback handler (unstaged).

When branched from a non-default branch:

Branch: feat/oauth-refresh (2 commits ahead of develop, no uncommitted changes)

Active intent: Implement token refresh for OAuth providers.
Constraint: Must stay compatible with the session store changes on develop.

Priority order:

  1. Active intent (what we're building and why)
  2. Current approach (decisions made)
  3. Rejected approaches (what NOT to re-explore — critical)
  4. Constraints (hard boundaries)
  5. Learnings (things that save time)
  6. In-progress work (unstaged/staged changes)

If intent evolved during the branch (a pivot), show both the original and current intent to make the pivot visible.

For Scenario B (branch with no commits):
Branch: feat/new-feature (0 commits ahead of develop)

No contextual history on this branch yet.
Staged: 2 files (src/auth/provider.ts, src/auth/types.ts)
Unstaged: none

Recent project activity (from develop):
  - Auth: OAuth provider framework merged, Google working
  - Payments: Multi-currency support shipped (EUR, GBP alongside USD)
For Scenario C (default branch, no changes):

Synthesize recent merged work from the last 20 commits. Group by area of activity. Surface any active constraints or learnings that apply broadly.

Recent project activity:

Auth: OAuth provider framework merged. Google working, GitHub and Apple planned.
  - Rejected auth0-sdk (session model incompatible with redis store)
  - Constraint: redis session TTL 24h, tokens must refresh within window

Payments: Multi-currency support shipped (USD, EUR, GBP).
  - Per-transaction currency, not account-level
  - Constraint: Stripe locks currency at PaymentIntent creation
  - Learned: presentment ≠ settlement currency in Stripe

What do you want to work on?
For Scenario D (default branch with uncommitted changes):

Same as Scenario C, with uncommitted changes noted at the top.

When there are no contextual commits at all

If the history contains conventional commits but no contextual action lines, still produce useful output from what exists:

Branch: main (no contextual commits found in recent history)

Recent activity (from commit subjects):
  - auth: 6 commits (OAuth implementation, middleware updates) — 2 days ago
  - payments: 4 commits (currency handling, Stripe integration) — last week
  - tests: 3 commits — last week
  - config: 2 commits

Most recent: feat(auth): implement Google OAuth provider

No explicit intent, decisions, or constraints are captured in commit history.
Using the contextual-commit skill on future commits will surface
the reasoning behind changes that commit subjects alone cannot show.

This is honest about the signal quality while still providing useful orientation. The suggestion to adopt the skill is a natural next step, not a sales pitch.

Show full SKILL.md (351 more words)Show less
Guidelines
  • Dense over conversational. Every line should carry information. No "Here's what's been happening" or "Let me tell you about."
  • Grounded in data. Only report what the action lines, commit subjects, and diffs actually show. Do not infer, speculate, or fill gaps.
  • Surface rejections prominently. Rejected approaches are the highest-value signal — they prevent wasted exploration.
  • Group by scope when multiple scopes exist. On the default branch with broad history, organize by domain area rather than chronologically.
  • End with a prompt. Close with "What do you want to work on?" or similar. Keep it short.
  • Scale to the data. If there are 2 contextual commits, the output is 3-4 lines. If there are 20, it's a few grouped paragraphs. Never pad thin data into a long output.

Scope Query (recall <scope>)

Targeted query across full repo history for a given scope. Prefix matching: auth matches auth, auth-tokens, auth-library, etc.

Step 1: Query
bash
git log --all --grep="(${SCOPE}" --format="%H%n%s%n%b%n---COMMIT_END---"
Step 2: Extract

From the gathered commit bodies, extract lines matching:

bash
grep -E "^(intent|decision|rejected|constraint|learned)\(${SCOPE}"

This captures all action lines whose scope starts with the query term.

Step 3: Output

Group by action type, chronological within each group. Show which sub-scopes were found.

Example (recall auth):

Scope: auth (also found: auth-tokens, auth-library)

intent:
  - social login starting with Google, then GitHub and Apple
  - original session approach incompatible with redis cluster setup

decision:
  - passport.js over auth0-sdk for multi-provider flexibility
  - JWT with short expiry + refresh token pattern

rejected:
  - auth0-sdk — locks into their session model, incompatible with redis store
  - auth-sessions — redis cluster doesn't support session stickiness needed by passport sessions

constraint:
  - callback routes must follow /api/auth/callback/:provider pattern per existing convention
  - redis 24h TTL means tokens must refresh within that window

learned:
  - passport-google needs explicit offline_access scope for refresh tokens
  - redis-cluster session affinity requires sticky sessions at load balancer level — too invasive

If no matches are found, say so plainly and suggest checking the scope name.


Action+Scope Query (recall <action>(<scope>))

Query a specific action type for a scope across full repo history.

Step 1: Query
bash
git log --all --grep="${ACTION}(${SCOPE}" --format="%H%n%s%n%b%n---COMMIT_END---"
Step 2: Extract
bash
grep "^${ACTION}(${SCOPE}"
Step 3: Output

Flat chronological list with commit subject for provenance.

Example (recall rejected(auth)):

rejected(auth) across history:

  - auth0-sdk — locks into their session model, incompatible with redis store
    from: feat(auth): implement Google OAuth provider

  - auth-sessions — redis cluster doesn't support session stickiness needed by passport sessions
    from: refactor(auth): switch from session-based to JWT tokens

If no matches are found, say so plainly.


Proactive Usage

Before making a significant decision in any scope area, check rejected and constraint lines for that scope first:

bash
git log --all --grep="rejected(${SCOPE}" --format="%b" | grep "^rejected(${SCOPE}"
git log --all --grep="constraint(${SCOPE}" --format="%b" | grep "^constraint(${SCOPE}"

If a previous commit recorded a rejected line for the approach you're about to propose, surface the rejection reason to the user before re-proposing it. The rejection may still stand, or circumstances may have changed — but the user should make that call with full context, not rediscover the same dead end.

Similarly, check constraint lines before proposing an approach that might violate a known boundary.

This check is lightweight and should become habitual for any scope you're actively working in.

© berserkdisruptors, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in skills/recall of berserkdisruptors/contextual-commits.

Open the folder on GitHubat commit 12efb1d

Compare with similar skills

Recall next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Recall compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Recall this skillberserkdisruptors/contextual-commits154—~2.8kAutomated safety check: PassMIT
ZCF Release AutomationUfoMiao/zcf6.1k—~3.4kAutomated safety check: PassMIT
Openqodexopenqodex/openqodex573—~2.5kAutomated safety check: PassApache-2.0
Pypi ReleasealchemiststudiosDOTai/tunacode125—~2.2kAutomated safety check: PassMIT
Update Dependenciesalorence/django-modern-rpc111—~1.3kAutomated safety check: PassMIT
npm Package Publisherklaudworks/universal-skills181—~923Automated safety check: PassMIT

Similar skills

  • Automates a version release with changesets: analyzes code changes, writes a bilingual CHANGELOG, bumps the version and commits through a release branch and pull request.

    6.1k GitHub stars~3.4k tokensUpdated 1 mo ago
    DevelopmentAuto-check passed
  • Openqodex

    openqodex/openqodex

    Code review for the current change, before it is pushed. An agent skill from openqodex/openqodex.

    573 GitHub stars~2.5k tokensUpdated today
    DevelopmentAuto-check passed
  • Pypi Release

    alchemiststudiosDOTai/tunacode

    This skill should be used when releasing tunacode-cli to PyPI.

    125 GitHub stars~2.2k tokensUpdated 5 days ago
    DevelopmentAuto-check passed
  • Update Dependencies

    alorence/django-modern-rpc

    Routine update of all project dependencies — uv itself, uv.lock (all groups), tool versions pinned in GitHub workflows and .pre-commit-config.yaml (uv, ruff, mypy...), and SHA-pinned GitHub Actions.

    111 GitHub stars~1.3k tokensUpdated 3 days ago
    DevelopmentAuto-check passed
  • npm Package Publisher

    klaudworks/universal-skills

    Releases an npm package by committing changes, bumping the version with npm version, pushing the tag and checking the GitHub Actions publish.

    181 GitHub stars~923 tokensUpdated 9 mo ago
    DevelopmentAuto-check passed
  • Kt Search Release

    jillesvangurp/kt-search

    A skill your agent uses when the user wants to cut, publish, tag, or create a GitHub release for kt-search, especially when the task includes version bumping, validating that commits are pushed…

    155 GitHub stars~1.2k tokensUpdated 3 mo ago
    DevelopmentAuto-check passed

Categories

Questions about Recall

What does Recall do?

Reconstruct and narrate the current development context from contextual commits. Recall is an agent skill from berserkdisruptors/contextual-commits. Reconstruct and narrate the current development context from contextual commits.

When should I use Recall?

Recall fits situations like: development work in your project.

How do I install Recall in Claude Code?

Run `npx skills add berserkdisruptors/contextual-commits --skill recall -a claude-code`. Or copy the skill folder (skills/recall in berserkdisruptors/contextual-commits) into .claude/skills/recall in your project. Claude Code loads it when a task matches its description.

How do I install Recall in Codex?

Run `npx skills add berserkdisruptors/contextual-commits --skill recall -a codex`. Or copy the skill folder (skills/recall in berserkdisruptors/contextual-commits) into .agents/skills/recall in your project. Codex loads it when a task matches its description.

Can I use Recall in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add berserkdisruptors/contextual-commits --skill recall -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/recall, .gemini/skills/recall, .github/skills/recall and .opencode/skills/recall in your project.

What does Recall need to run?

Going by SKILL.md and its folder, Recall needs the command-line tools its instructions call (git).

Does Recall access the network?

SKILL.md contains no URLs. Its commands use git, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Recall safe to install?

Our automated static check of SKILL.md found no risky patterns, such as piping downloads into a shell, reading credential files or hidden Unicode. It is not a guarantee. Review the folder before installing.

What licence does Recall use?

Recall is published under the MIT licence (declared in SKILL.md). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Recall use?

About 2.8k tokens (SKILL.md is roughly 11k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Recall?

Skills that share tags, products or a category with Recall: ZCF Release Automation (UfoMiao/zcf, 6.1k stars), Openqodex (openqodex/openqodex, 573 stars), Pypi Release (alchemiststudiosDOTai/tunacode, 125 stars) and Update Dependencies (alorence/django-modern-rpc, 111 stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Recall?

berserkdisruptors (a GitHub organization) maintains it in berserkdisruptors/contextual-commits, which has 154 GitHub stars. The repository was last updated on May 7, 2026.

Source: berserkdisruptors/contextual-commits on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.