Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/jonny981/claude-tandem/commitnpx skills add jonny981/claude-tandem --skill commitgit clone --depth 1 https://github.com/jonny981/claude-tandemWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00039 | $0.00914 |
| Opus 5 | $0.00019 | $0.00457 |
| Sonnet 5 | $0.00008 | $0.00183 |
| Haiku 4.5 | $0.00004 | $0.00091 |
Grade A, and why
commit scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 63 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Tandem Commit
Orchestrate a structured commit and run the full memory pipeline. The commit body is the primary, permanent epistemic record. After the commit, progress.md is compacted into MEMORY.md and reset.
Prerequisites
Files must already be staged (git add). If nothing is staged, tell the user and stop.
Step 1: Gate on progress.md
Read progress.md from .claude/progress.md at the repo root (or ~/.tandem/progress/<slug>/ if not in a git repo).
- If progress.md is missing or template-only (Working State fields are still placeholders): STOP. Write your Working State first (current task, approach, blockers, key files). Then continue. A commit without session context produces a weak epistemic record.
- If progress.md has content: proceed.
Step 2: Read context
- Read progress.md (full session state)
- Scan your auto-memory directory for
feedback_*.mdfiles. Read any whose description suggests relevance to the staged changes. These capture user corrections and preferences that shaped the approach, exactly the kind of context the commit body should preserve. - Run
git diff --cached --statto see what is staged - Run
git diff --cachedfor the full diff
Step 3: Author the commit
Write the commit following the conventions in tandem-commits.md (the rules file, not this skill). The rules file defines the subject format and body structure. This skill orchestrates, the rule defines format.
Key principles from the rules:
- The subject uses Conventional Commits format
- The body is the permanent epistemic record, structured with section headers
- Write for machine comprehension: explicit intent, reasoning, epistemic state
- The diff shows what changed; the body captures what the diff cannot
Condensing progress.md into the commit body: The session log is raw material, not the final product. Apply reasoning when distilling it:
- Include only what is relevant to the staged changes. Tangents, unrelated investigations, and dead-end explorations that didn't affect the final code should be omitted.
- Identify the coherent narrative: what problem was being solved, what iterations occurred, what the user's corrections were, what constraints shaped the final design.
- If the session covered multiple unrelated topics, only capture the thread that produced the staged changes.
- Prioritise reasoning that would be hard to reconstruct from the diff alone: the why behind a design choice, constraints that aren't visible in code, alternatives that were rejected and why.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 63 lines · 39 tokens per session scan A d158384cd633
commit is a skill published in the GitHub repository jonny981/claude-tandem (6 stars, last pushed 4mo ago), licensed MIT. It adds 39 tokens to every session and 914 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other skills, from other repositories
commit
Stage and commit changes using conventional commits format.
git-commit
This skill should be used BEFORE running any git commit command. Triggers when about to run git commit. Ensures commit messages follow Conventional Commits specification and prompts for the Jira ticket number.
commit
Stage the appropriate changes and commit them with a Conventional Commits message.
commit
A commit candidate is the exact patch that will enter the commit. A coherent change has one task or issue as its reason for existing.
git-commit
Creates git commits following Conventional Commits format with type/scope/subject. Use when user wants to commit changes, create commit, save work, or stage and commit. Enforces project-specific conventions from CLAUDE.md.
git-workflow
Defines branch naming conventions, PR template requirements, commit message format, discovered-issues escalation policy, task tracking conventions. Load when committing, pushing, or opening PRs.