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 commands/codeverbojan/claude-code-kickstart/retrospectivegit clone --depth 1 https://github.com/codeverbojan/claude-code-kickstartWhat 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 | $0.00016 | $0.00693 |
| Opus 5 | $0.00008 | $0.00347 |
| Sonnet 5 | $0.00003 | $0.00139 |
| Haiku 4.5 | $0.00002 | $0.00069 |
Grade A, and why
retrospective 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 — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Retrospective — Learn From Mistakes
Analyze .claude/signals.jsonl and generate new rules for gotchas.md.
1. Read Signals
Read .claude/signals.jsonl. Each line is a JSON object with:
timestamp: when the signal was capturedtype:revert,test_failure,lint_failure,typecheck_failuredetail: human-readable descriptioncommand: the command that triggered the signal (secrets redacted)snippet(optional): first 500 chars of failure output — use this for root cause analysis
If the file doesn't exist or is empty, report "No signals captured yet. Work on the project and mistakes will be automatically detected."
2. Group and Analyze
Group signals by type and look for patterns:
- Repeated reverts on the same file → something about that file's patterns isn't understood
- Repeated test failures → a testing pattern or convention is being missed
- Repeated lint failures → a linting rule keeps being violated
- Repeated typecheck failures → a type pattern isn't being followed
For each group, use the snippet field (when available) to identify the ROOT
CAUSE pattern. "Tests failed 5 times" is a symptom. "Form validation was
missing in 3 of 5 failures" is a pattern. If no snippet is available, group
by command pattern and flag which files/commands repeatedly trigger failures.
3. Read Existing Gotchas
Read gotchas.md to check what rules already exist. Do NOT create duplicate
rules. If an existing rule covers the pattern, skip it.
4. Generate New Rules
For each new pattern found, write a gotcha rule. Format:
N. [CATEGORY] Rule text — learned from [X signals over Y days]
Categories: [REVERT], [TEST], [LINT], [TYPECHECK]
Rules should be:
- Actionable — tell Claude what TO DO, not just what went wrong
- Specific — reference actual file patterns, commands, or conventions
- Preventive — address the root cause, not the symptom
5. Append to gotchas.md
Append new rules under a ## Auto-generated section at the end of gotchas.md.
If this section doesn't exist, create it. Find the highest existing rule number
in the file (manual or auto-generated) and continue from there. Never overwrite
or reorder existing rules.
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 · 80 lines · 16 tokens per session scan A e6fb978befd2
retrospective is a command published in the GitHub repository codeverbojan/claude-code-kickstart (2 stars, last pushed 4mo ago), licensed MIT. It adds 16 tokens to every session and 693 once invoked, about $0.0001 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 commands, from other repositories
stats
Show token-saver compression statistics and savings.
implement
Execute implementation by processing atomic task files one at a time with Context Pinning (Atomic Traceability Model).
cleanup
Detect and remove orphaned code, unused components, dead routes, and stale database artifacts.
specify
Create or update the feature specification from a natural language feature description.
_registry-protocol
This protocol is MANDATORY for ALL commands, agents, and phases.
amby.converge
Check the codebase against spec, plan, and tasks after implementing; append gap tasks, never edit code.