Skip to content

sprngr/rubber-duck / skills/duck-tape

duck-tape

Verified

Two-tier session memory: CONTEXT.md (persistent) + .duck-tape working state, with merge/dedupe. Subcommands: merge, resume, init, prune, migrate. Use when: "duck-tape", "compact session", "update CONTEXT.md", "resume session".

FileKindSize
README.mdmarkdown15.5 KB
SKILL.mdskill md20.7 KB
examples/CHANGELOG.mdmarkdown419 B
examples/CONTEXT.mdmarkdown2.1 KB
examples/STATE.mdmarkdown1.2 KB
examples/bootstrap-CONTEXT.mdmarkdown463 B
hooks/claude-code.hooks.jsonconfig291 B
hooks/claude-code.hooks.windows.jsonconfig312 B
hooks/copilot.hooks.jsonconfig325 B
hooks/extract-raw.ps1script15.2 KB
hooks/extract-raw.shexecutablescript12.2 KB
hooks/extract-state.ps1script15.1 KB
hooks/extract-state.shexecutablescript11.3 KB
hooks/opencode.plugin.jsscript12.9 KB
hooks/pre-compact.ps1script1.1 KB
hooks/pre-compact.shexecutablescript1.0 KB
references/GUARDRAILS.mdmarkdown757 B
references/HOOKS_GUIDE.mdmarkdown11.3 KB
references/OUTPUT_SCHEMA.mdmarkdown3.8 KB
references/SCHEMA.mdmarkdown7.0 KB
references/STATE_SCHEMA.mdmarkdown5.7 KB

references/STATE_SCHEMA.md

View on GitHub
# Session State Schema (Tier 2)

## File

Path: `.duck-tape/<session_id>.state.md`
Session ID: `<YYYY-MM-DD-HHMM>` (timestamp-based, generated by duck-tape)

## Agent State Schema

```
# Agent State

Session: `<session-id>` | Cwd: `<absolute-path>` | Repo/Branch: `<repo>@<branch>`
Created: `<ISO-8601>`

## Approved Workflow
<what was approved in user's terms. Link spec/plan/ticket if one grounds this work>

## Position
- **Current:** <one step in flight>
- **Done:** <ordered newest last>
- **Remaining:** <in order>

## Decision Log
<ISO-8601> APPROVED/REJECTED: <what> - <reasoning>
<one entry per decision, append new>

## Established Facts
<facts established during session>

## Re-derivation
<how to recover state not written above: which commands to run, which files to read, which endpoint to query. Assume reader has lost all conversation history>

## Suggested Skills
<optional. Skills next session should invoke. Consumed by next agent, does not translate to CONTEXT.md>
```

## Section Definitions

### Header

Metadata block. Fields:

- `Session`: session ID (`<YYYY-MM-DD-HHMM>`)
- `Cwd`: absolute path of working directory
- `Repo/Branch`: `<repo>@<branch>` from git
- `Created`: ISO-8601 timestamp of first write

### Approved Workflow

What was approved in the user's terms. Link spec/plan/ticket if one grounds this work.

Translation: one key per approved workflow -> CONTEXT.md **Decisions**. Verbatim text.

### Position

Current state of work.

- `Current`: one step in flight.
- `Done`: ordered newest last.
- `Remaining`: in order.

Translation:

- `Current` -> state-file-local. Not translated to CONTEXT.md. Next agent reads state file on resume.
- `Done` -> state-file-local. Not translated to CONTEXT.md. Next agent reads state file on resume.
- `Remaining` -> CONTEXT.md **Open-Questions**, one question per item.

### Decision Log

Append-only. One entry per decision.

Format: `<ISO-8601> APPROVED/REJECTED: <what> - <reasoning>`

Translation: one key per decision -> CONTEXT.md **Decisions**. Status (APPROVED/REJECTED) preserved in entry text.

### Established Facts

Facts established during session.

Translation: one term per fact -> CONTEXT.md **Glossary**. Verbatim.

### Re-derivation

Recovery instructions for lost conversation history. Commands to run, files to read, endpoints to query. Assume reader has lost all conversation history.

Translation: none. Consumed by next agent directly from state file. Does not land in CONTEXT.md. Same category as Suggested Skills: state-file-local, not translated to persistent context.

### Suggested Skills

Optional. Skills next session should invoke.

Translation: none. Consumed by next agent directly from state file. Does not land in CONTEXT.md.

## Translation Map (Rigid)

| State section | CONTEXT.md target | Rule |
| --- | --- | --- |
| Approved Workflow | Decisions | one key per approved workflow, verbatim text |
| Position.Current | (none) | state-file-local, not translated |
| Position.Done | (none) | state-file-local, not translated |
| Position.Remaining | Open-Questions | one question per remaining item |
| Decision Log | Decisions | one key per decision, status preserved in text |
| Established Facts | Glossary | one term per fact, verbatim |
| Suggested Skills | (none) | consumed by next agent, not translated |
| Re-derivation | (none) | consumed by next agent, not translated. State-file-local. |

No interpretation. Each state section maps to exactly one CONTEXT.md target. Re-derivation and Suggested Skills are state-file-local: not translated. Deterministic.

## Translation boundary

State files retain session detail needed for recovery, including rejected
approaches, review or test references, temporary approval flags, commands, and
position history. Do not remove that detail from Tier 2.

Before writing CONTEXT.md, remove session narration and temporary markers that
only explain how work was produced. Preserve the underlying durable decision,
constraint, convention, fact, debt, or open question when one exists. Keep a
test or review reference only when it states a durable verification requirement.
Keep historical rationale in an ADR or changelog, not ordinary CONTEXT.md.

Example:

State entry:

```text
APPROVED: use shared comment guidance - review rejected duplicated wording
```

Durable CONTEXT.md entry:

```text
- **Comment guidance**: use shared comment guidance.
```

## Lifecycle

### Write

Single write on `/duck-tape` invocation. No mid-session checkpoints.

### Rotation

- Max 10 state files in `.duck-tape/`.
- Eviction precedence: auto dropped first (oldest auto), then recovered, then manual.

### gitignore

- Auto-create `.duck-tape/.gitignore` with content `*` on first write if missing.
- Self-contained. Does not touch root `.gitignore`.
- State files not meant for commit.

## Redaction

Applies to all state content before write. Same rules as CONTEXT.md redaction:

- Scan for secrets/PII: API keys, passwords, tokens, connection strings, env var values, personally identifiable information.
- On detection: reject flagged content. Report findings. Ask user to redact source or confirm mask-in-place (`<REDACTED>`).
- If a state-file write would contain raw secrets, reject it; write masked content only after user confirmation.
- Redaction applies to all sections including Re-derivation and Suggested Skills.

## Boundaries

- State file uses Agent State schema only. No CONTEXT.md sections in state file.
- Translation is rigid. No interpretation of content to pick target section.
- Re-derivation is verbatim in state file. If content is re-derivation, keep it verbatim; do not rewrite or summarize. Not translated to CONTEXT.md.
- Suggested Skills does not translate. Next agent reads state file directly.
- State files are working memory, not persistent. Rotation cap enforced.
slack-gif-creatoranthropics/skillsKnowledge and utilities for creating animated GIFs optimized for Slack. Provides constraints, validation tools, and animation concepts. Use when users request animated GIFs for Slack like "make me a GIF of X doing Y for Slack."Stars: 198kUpdated Claude CodeCodexVerifiedInstall
ponytailDietrichGebert/ponytailLazy senior dev mode for any coding task: the smallest change that fully solves it, and a reply a busy human understands in one read.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailHarvest every shortcut: comment into one debt ledger, so deferrals get tracked instead of forgotten. One-shot report.Stars: 196kUpdated Claude CodeCodexVerifiedInstall
ponytail-debtDietrichGebert/ponytailList every `shortcut:` comment (and older `ponytail:` ones) as a debt ledger. One-shot report, changes nothing. Use for "ponytail debt", "what did ponytail defer", "list the shortcuts", /ponytail-debt.Stars: 196kUpdated Claude CodeCodexVerifiedInstall